Hulu + Live TV lags slightly behind at 4.6 million subscribers, according to MoffettNathanson.

YouTube TV Tops Other Live TV Streaming Services With 6.5 Million Subscribers | PCMag

Suivez en direct toute l'actualité politique et économique : élections, finances publiques, croissance et fiscalité sur capital.fr

le pape du crudivorisme» récolte près de 175 000 euros

Найти в сети мультики для детей на украинском языке не так уж сложно. Для этого можно, в частности, воспользоваться YouTube - популрным видеохостинго...

Лучшие детские YouTube-каналы с мультиками на украинском - список | РБК Украина

Après la première vidéo du youtubeur Cyprien annonçant sa collaboration avec la Ville de Nice pour le prochain Carnaval de Nice, celui-ci a dévoilé dans une vidéo sur les réseaux sociaux le dessin qu’il a imaginé pour un...

Carnaval de Nice 2024. Cyprien, star de YouTube, dévoile le dessin de son char

Тимчасове блокування популярного в Білорусі відеохостингу YouTube відбулося перед вітальним словом лідерки демократів Світлани Тихановської і тривало близько ...

Як у Білорусі блокували YouTube і чи можна це зробити повністю

Переяславський діджей та хореограф, а віднедавна й військовослужбовець Євгеній Лой став учасником нового проєкту «Історія за бабки» на українському youtube-каналі

Військовий з Переяслава в youtube-шоу «Історія за бабки» розповів про те, як втрапив у пікантний вебкам-шантаж (відео)

Савеловский суд Москвы отказался удовлетворить иск предпринимателя и общественного деятеля Александра Морозова заблокировать YouTube-канал ведущей и блогера ...

Суд решил не блокировать YouTube-канал Ивлеевой из-за видео с нецензурной лексикой

Ιδανικά κλείνει το 2023 για το Gangnam Style, καθώς το βίντεο κλιπ του συνεχίζει να κάνει θραύση στο YouTube, με το νέο έτος να το βρίσκει να σπάσει ένα τεράστιο ρεκόρ, ...

Το Gangnam Style έσπασε ένα μεγάλο ρεκόρ του YouTube λίγο πριν μας αποχαιρετήσει το 2023 - Newsbeast

A YOUTUBER was found dead by his girlfriend at his home in Brazil as police investigated his alleged history of sharing child abuse images. Paulo Cezar Goulart Siqueira, known as PC Siqueira to his…

YouTube star PC Siqueira with 2million subscribers found dead by his girlfriend while facing paedo cop probe | The Sun

De precursor do YouTube a investigado: relembre a trajetória de PC Siqueira

De precursor do YouTube a investigado: relembre a trajetória de PC Siqueira | CNN Brasil

Yoto

yoto

ANGEL VISITS.

From The Jetsons to Juicero: making the world a more comfortable place For the last decade, technology promised to ‘ ’. But is it delivering? Let's take a look at some products and services that received substantial media attention and funding since the early 2010s to find out: make the world a better place allowed users to send the word "Yo" to their contacts. Yo , then Stockwell, aimed to replace corner stores by installing vending machines in residential buildings (ideal for convenient toilet paper replacement. Those machines spared me some pretty desperate convenience store runs when I lived in San Francisco) Bodega sold a $400 juicer that was essentially a glorified bag-squeezer. Juicero developed a wearable device that claimed to induce relaxation or stimulation through electrical signals sent to the brain. Thync created a solar-powered inflatable lantern. LuminAID is a wearable device that delivers electric shocks to users when they engage in "bad habits" like biting their nails or oversleeping. Pavlok , then Spoonr, allowed users to find people in their area to cuddle with Cuddlr allowed people to rent out their bathrooms to strangers for a fee. Airpnp is a dating app that matches people based on their shared dislikes. Hater promised to create a bracelet that would project a touchscreen onto your skin, allowing you to control your phone without actually touching it. Cicret Amidst the deluge of tech discoveries and inventions promising utopia, the reality remains underwhelming. The majority of the list addresses niche needs. For many entrepreneurs, the tongue-in-cheek aim is to supply everything their college campuses (or mothers) no longer provide—an infantilized version of The Jetsons' future. When everything is dubbed "world-changing," what truly is? Clay Tarver, writer and producer of HBO's satirical "Silicon Valley," notes some big tech companies have forbidden employees from saying "We're making the world a better place." Tarver jests, "At the very least, we're making the world a better place by making these people stop saying they're making the world a better place." If you think the list of past flops might be biased, you may have a point. However, hindsight is always twenty-twenty. In reality, even recent tech trends are rife with startups suffering from a Napoleon Complex. Economist Robert J. Gordon contends that today's AI, which focuses on pattern recognition, isn't as groundbreaking as electricity or the internal combustion engine, despite what technologists and investors claim. Don't expect a world-shattering revolution to happen anytime soon. We're in a comfort crisis. The brightest minds in science and technology, from founders to influential investors, seem preoccupied with hype, external validation, and conformity. Prioritizing venture capital over sustainable businesses, chasing trends heedless of unique needs, and surrounding oneself with "yes" people rather than seeking critical feedback and diverse perspectives are all misguided endeavors. Countless talented individuals opt for safe, incremental ventures, such as SaaS companies and cookie-cutter consumer startups. Although these pursuits may yield some benefits, they ultimately neglect crucial human challenges, including existential ones. This not only stifles progress but also represents a missed opportunity for gifted individuals to leave a lasting impact on the world. In essence, for any exceptionally capable person, shirking a noble quest isn't just neutral—it's a net negative for society. Velocity is everything A worthy endeavor should target a problem substantial enough to rightfully claim social impact credentials. Tough challenges with a low chance of success, yet a social promise enticing enough to overlook operational complexities—think artificial general intelligence (AGI), curing Alzheimer's disease, sustainable nuclear fusion, or even building a . network state However, just because the ‘mission’ is ‘impossible’, it doesn’t mean it’s worth pursuing – like “ ”, whatever all of that means. The difficulty axis is just one piece of the puzzle and a somewhat one-dimensional perspective on problems. Hard missions without an angle are an exercise of style at best, like the , and pure fugazi at worst, like Theranos. shaping the future of web3 through storytelling, experiences, and community war on cancer The need for ambitious endeavors to improve the world is driven by technological velocity, which is often measured by the degree of operating leverage - the ratio of fixed costs to variable costs, that explains how much of a company’s operating income changes in response to a change in revenues. Unit economics tells the true story of broad adoption. It’s either that or knowing a special secret, a glimpse of truth at a technological level that can flip the game on its head. Moonshot builders can be grouped into two types: those with in-depth knowledge and understanding of technology, research, and fieldwork, and those who excel in execution and operations. The first group discovers bold breakthroughs, while the second accelerates development and ensures they become a reality. Both roles are critical in meeting the needs of society, and it's important to recognize one's unique strengths within this framework. Your author is an action-taker. As a non-scientific, non-technical expert, I have very little to say about the process of scientific insight generation. But I am interested in both very large and very small patterns – how markets and human behavior change, as well as how the unit economics of a product, at its most atom-like level, can narrate the story of technological velocity. When it comes to resource allocation – whether that is money, time, or energy – velocity is everything. Three case studies Exceptional quests are those that address significant societal issues while showcasing remarkable velocity. While assessing a quest's potential impact might be simple, understanding velocity is more complex. Velocity represents a learning factor, indicating the pace of change. It serves as a proxy for decreasing adoption costs, which are crucial for turning breakthrough technologies into widely adopted products. The relationship between experience – measured as the cumulative installed capacity of the technology – and the price of that technology is called the learning curve of that technology. The relative price decline associated with each doubling of experience is the learning rate of a technology. That more production leads to falling prices are not surprising – such ‘economies of scale’ are found in many corners of manufacturing. If you are already making dinner, it isn’t that much extra work to accommodate an extra guest. For over six decades, the concept of technological learning has suggested that a technology's performance enhances as experience accumulates. I argue that the best way to anticipate all future technological advancements is to concentrate on paradigms that exhibit learning curves. There’s no shortage of examples of innovations that benefitted from technological learning curves. LED lighting has emerged as an affordable, energy-efficient alternative to traditional bulbs. Electric vehicles now offer a price-competitive substitute for gasoline cars. To get our expectations for the future right we ought to pay a lot of attention to those technologies that follow learning curves. Initially, we might only find them on a high-tech satellite out in space, but the future belongs to them. Most technologies obviously do not follow a learning curve – the prices of bicycles, fridges, or coal power plants do not decline exponentially as we produce more of them. But those which do – like computers, solar PV, and batteries – are the ones to look out for. They might initially only be found in very niche applications, but a few decades later they are everywhere. To get our expectations about the future right, I argue that today there are three important areas where upstream technological paradigms are exhibiting learning curves in their unit economics – semiconductors, cancer detection, and solar batteries. Moore’s Law and Semiconductors In the 1960s, Intel co-founder Gordon Moore observed that the number of transistors in an integrated circuit doubled at a clock-like pace, leading to greater processing power at the same cost. To Moore, there was no reason that progress couldn’t continue. And there hasn’t been. As semiconductors became cheaper and more efficient, they penetrated increasingly more aspects of our daily lives. By and large, they continued to drive improvements in things we already use, from phones to cars to appliances. And of course, there’s , which you’d have to be living under a rock at this point to not have heard of. Ask it anything and it’ll very confidently tell you… Well, something. chatGPT Morris Chang, known as the Godfather of Taiwan's chip industry, developed this concept into a revolutionary pricing model called 'learning curve pricing' or 'experience curve pricing,' which involved setting chip prices below initial costs to optimize production volume. As a result, he captured market share, enabling production lines to run at maximum capacity and reduce time to increase yields. While superlatives cannot adequately describe the profound effects of the semiconductor industry on human progress, the expanding use of semiconductors is a clear sign of the fundamental importance of this industry that stood the test of time. The minuscule transistorized computer chips made from semiconductors have become the lifeblood of our 21st-century information society, just as oil was to the industrial world of the 20th century. Without the ubiquitous electronic infrastructure powered by computer chips today, virtually nothing would run. Chip manufacturing may be the most complex, process-controlled industry the world has ever seen. Moore’s Law accurately charted the pathway in size reduction and thus in performance for those transistors. Investors, producers, system manufacturers, and consumers alike all benefited from the Swiss watch-like regularity of the computer chip product cycles. Unlike no other business in the world, the semiconductor industry had an accurate, proven roadmap for its future…with clear technological velocity, orchestrated to the rhythm of Moore’s Law. Flatley’s Law and early cancer detection Cancer continues to be a pressing public health issue, with mortality rates decreasing by merely 19% since 1990, compared to a decline of over 50% for cardiovascular diseases. The COVID-19 pandemic has further exacerbated the urgency to address cancer due to missed screenings and diagnoses affecting millions. Nonetheless, advancements in diagnostic and therapeutic innovations offer the potential to significantly reduce cancer mortality rates. Early detection is paramount, as cancer is a progressive disease, and advanced tumors account for a disproportionate number of deaths. Molecular testing is essential for precision therapy because it identifies tumor-specific mutations that guide oncologists in selecting the most effective treatments. Approximately two decades ago, when the human genome was first sequenced, it took a team of researchers at the National Institutes of Health (NIH) nearly 15 years and over $2.7 billion to accomplish the task. Today, a human genome can be sequenced in roughly a day for around $500, and the cost is anticipated to soon drop to $100. Flatley's Law (named after former Illumina’s CEO), an analog to Moore's, demonstrates even faster advances and more significant cost reductions. Flatley's Law highlights the declining cost of cancer detection tests, such as multi-cancer earlier detection (MCED) tests, which can identify multiple cancer types from a single blood draw. These tests have the potential to decrease cancer mortality rates by 15% at a $500 reimbursement price, marking the real turning point for widespread adoption. Reimbursable tests are the Schelling point of cancer detection. With falling prices and continuous improvement in accuracy, multi-cancer earlier detection (MCED) can become a part of a routine check-up for most and avert an unprecedented number of cancer-related deaths. Wright’s Law and solar batteries To reduce emissions the world needs to rapidly transition towards a low-carbon energy system. Around three-quarters of come from energy and industry. One of the barriers to this energy transition has been the relative cost of different energy sources. Fossil fuels were cheaper than renewables and therefore became the dominating sources of energy. global greenhouse gas emissions Thankfully this is changing quickly. The cost of renewable technologies has plummeted – they’re now cost-competitive or cheaper than new fossil fuels. In 2009, it was more than three times as expensive as coal. Now the script has flipped and a new solar plant is almost three times cheaper than a new coal one. The price of electricity from solar . declined by 89% between 2009 and 2019 But the cost of electricity technologies themselves is only part of what matters for this transition. One of the challenges that renewables face is that they produce energy intermittently. The sun doesn’t always shine and the wind doesn’t always blow and so we don’t get a steady flow of generation throughout the day. An obvious solution is to store excess energy and then release it later. But to do so, we need lots of energy storage and this adds large costs to our energy system. The price of lithium-ion battery cells . A battery with a capacity of one kilowatt-hour that cost $7500 in 1991, was just $181 in 2018. That’s 41 times less. What’s promising is that prices are still falling steeply: the cost halved between 2014 and 2018. A halving in only four years. declined by 97% in the last three decades To put this in perspective: the popular electric car – which is also one of the most affordable models – has a 40 kWh battery. At our 2018 price, the battery costs around $7,300. Imagine trying to buy the same model in 1991: the battery alone would cost $300,000. Nissan Leaf What this means is that batteries have been getting smaller and lighter for any given electrical capacity. You might have noticed this yourself as your mobile phones were getting lighter and slimmer. This is a crucial technological improvement as one of the major drawbacks of some battery technologies is that they are heavy and this limits their use in a number of technologies that are still fossil fuel powered. Making innovation irresistible The history of technology is shaped by groundbreaking innovation reaching widespread adoption. But the vast majority of “world-changing” ventures functionally amount to economic noise. To achieve double-digit GDP, cure cancer and halt environmental decay we need a bold tech with straightforward economics. Bull markets, abundant venture capital funding, and the Zuckerberg-esque mythology of the rebel founder muddled the waters of the second half of this formula - celebrating outsized ambition not only as a necessary but also a sufficient condition. This is totally backwards. Moreover, one of the very worst misconceptions about technological paradigms is that they happen overnight. They don’t. Societal technological revolutions are exceedingly rare and the implementation challenges are much larger than the technical considerations alone. The chronicle of technological progress is interspersed with a history of setbacks. This is attributable to diseconomies of scale and the inability to achieve the necessary economic curves that lower prices and promote adoption. Nuclear power serves as a prime illustration of this phenomenon. In many places, building a power plant has become more expensive over time. This is of course very unfortunate since nuclear is both a low-carbon source of electricity and one of the safest sources of electricity as we have seen. One reason for rising prices is the increased regulation of nuclear power. A second reason is that the world has not built many nuclear power plants in recent years so supply chains are small, uncompetitive, and are not benefiting from economies of scale. Learning curves, after all, mean transferring the knowledge gathered in one instance to another. No repetition, no learning. Reducing the costs of innovative technological paradigms is crucial for achieving widespread adoption. The technical advantages of MCEDs compared to traditional biopsies or renewable energy over fossil fuels remain largely insignificant to society unless these innovations can achieve cost parity with their alternatives. The key indicators of progress lie in unit economics. If you are actively involved in, investing in, or simply passionate about the technology industry, I encourage you to take a step back and carefully evaluate your work at a 10-minute grocery delivery startup or a tool designed for Web3 communities.  Have a hard, critical look at their mission and consider whether you are building " " that just… squeezes giant ketchup sachets of fruit. an incredibly complicated piece of engineering Also published . here

Are You Building an Incredibly Complicated Piece of Engineering Just to Squeeze... Ketchup?

আপনি কি স্কুইজ করার জন্য ইঞ্জিনিয়ারিংয়ের একটি অবিশ্বাস্যভাবে জটিল অংশ তৈরি করছেন... কেচাপ?

Bauen Sie ein unglaublich kompliziertes Stück Technik, nur um ... Ketchup herauszupressen?

¿Estás construyendo una pieza de ingeniería increíblemente complicada que simplemente... exprime salsa de tomate gigante?

Êtes-vous en train de construire une pièce d'ingénierie incroyablement compliquée qui ne fait que… presser du ketchup géant ?

क्या आप इंजीनियरिंग के एक अविश्वसनीय रूप से जटिल टुकड़े का निर्माण कर रहे हैं जो सिर्फ ... विशालकाय केचप को निचोड़ता है?

巨大なケチャップを絞るだけの信じられないほど複雑なエンジニアリングを構築していますか?

케첩을 짜내기 위해 엄청나게 복잡한 엔지니어링 조각을 만들고 계시나요?

Você está construindo uma peça de engenharia incrivelmente complicada que apenas… aperta o ketchup gigante?

Sırf Ketçap Sıkmak İçin İnanılmaz Karmaşık Bir Mühendislik Parçası mı İnşa Ediyorsunuz?

Bạn đang xây dựng một bộ phận kỹ thuật cực kỳ phức tạp mà chỉ… vắt được nước sốt cà chua khổng lồ?

你是在建造一个极其复杂的工程，只是……挤巨型番茄酱吗？

Knitting as Programming or how I learned programmers owe it all to fiber As part of my bootcamp, I was asked to present for 10 minutes on any topic. I've been crocheting since 2013 and knitting since 2018, and I had noticed some similarities between the way I approached a project in either fiber arts or programming. What I did not expect to find was a rich history of fiber arts and programming influencing each other. The Beginning of the Relationship Between Fiber and Programming Fiber arts, like knitting, are crafts using yarn, string, thread, etc. This encompasses thousands of years of history, starting in 6500 BCE with Viking . Other crafts included under this umbrella are crochet, embroidery, weaving, and textile making. Nålbinding Historically, fiber arts and programming have been very intertwined, starting with the Jacquard Loom, an attachment for powered fabric looms. It was the first machine to use a chain of punch cards to automate instructions. It enabled the loom to make intricate patterns in textiles based on the punch cards it was fed. It inspired Charles Babbage in making his analytical machine. A century later, his designs and Ada Lovelace's notes were read by the people who built the first computer. The relationship between fiber arts and programming gets even more overt with (also called Little Old Lady Memory). Used in the 1960's by NASA, software written by MIT programmers was woven into core rope memory by female workers. Core Rope Memory The direction the current was forced through the wires induced field forces to circulate in a clockwise or counterclockwise direction. One direction is a stored 1, while the other is a stored 0. The result was non-volatile storage, which could retain a large amount of information without power - roughly 2.5 megabytes per cubic meter. This was an 18-fold improvement over magnetic-core memory, the standard solution used at the time. Present Day Like computers, the textile industry has come a long way since the Jacquard Loom, integrating more and more automated manufacturing. Perhaps you'd like to ? Makers will find a way, so when home knitting machines were no longer produced, . 3D print a lot of knit toys people built the system themselves was established in 2007 with free membership and only fiber arts ads. Today, Ravelry is the focus of the fiber arts community with over 9 million users. Their website was the first to allow people to exchange information, tools, materials, and patterns in pdf format on a large scale and in every language, revolutionizing the industry. Furthermore, they offer discussion groups, so people can get help with patterns and techniques. Ravelry Owned and operated by 6 women, they employ a tech stack of about 10 Gentoo Linux servers, MySQL 8.0, Ruby on Rails, Manticore Search, Amazon S3 for image storage and a CDN, Redis, memcached, nginx, and haproxy. By removing barriers to entry, Ravelry is credited with creating an explosion of interest in knitting and crochet, driving innovation in easier-to-follow patterns. I've personally talked with local yarn store owners about their experience before and after Ravelry, and they describe it like night and day. Knitters are now primarily 25–35 years old. Local, regional, and international knitting groups, fiber festivals, and yarn crawls have cropped up everywhere. Indie dyed yarn is now a huge business. There's renewed interest in fiber shows, which educate people on where their yarn comes from. The audience for blogs and podcasts related to fiber arts in general has grown exponentially. A Knitting Primer When you knit something, you use two knitting needles. All of your stitches will be live, meaning if you drop a stitch the fabric you're creating will unravel. Patterns are created by alternating types of stitches or yarn of different colors. For example, K2P2 would mean 2 knit stitches followed by 2 purl stitches. The difference between stitches is how you feed the new loop of yarn through the live stitch on your needle. Effectively, a knit stitch goes through the front of the loop. A purl stitch goes through the back of the loop. Repeating this for several rows is how you would get a ribbed brim on a hat. Patterns as Programming Knitting pattern designers follow DRY. Pattern repeats are for loops, typically surrounded by brackets, asterisks, or parentheses. It's common to see while loops like 'until the end of the round' or 'to n stitches before marker'. Higher-level languages like Python and Javascript use symbols and keywords to tell the computer what the desired output is, just like knitting charts. Like assembly languages, operands are pushed and popped onto the needles. Patterns are tested and retested by users for correctness, scalability, and the look of the finished object (aka expected output). Pattern developers prefer open-source software ( , , , , etc.). Darktable Gimp Libre Stitchmastery Different placements of yarn and needles produce different outputs. This shawl was created using 8 stitches - all knit based (no purl). One row was written like this: k3, SM, (k2tog, yo) 2 times, [k1, yo, ssk, yo, sk2p, yo, k2tog, yo] to 5 stitches before marker, k1, (yo, ssk) 2 times), SM, k3 There have even been two (no longer supported) programming languages designed specifically for hand knitting - KnitML, a markup language, and KEL, inspired by Groovy, a Java language. Patterns as Regular Expressions Say we want to make a square of ribbing. is the for one knit stitch followed immediately by one purl stitch. Curly brackets are used for repeats, so repeated 10 times looks like this: . For a new row we use a new line, written, . The full regular expression for the row is then . Since we turn the square to knit the next row, the following line must be s, or . We want our square to be 40 rows, so they are in turn wrapped up in yet more brackets, producing: . A more general pattern would look like . kp regular expression kp (kp){10} \n (kp){10}\n pk (pk){10}\n ((kp){10}\n(pk){10}\n){20} ((kp){x}\n(pk){x}\n){y} Patterns as Tools for Teaching Programming "Knitting patterns have evolved from vague, chatty discourse written for experts to precise, line-by-line procedures that are akin to programs." -- " " by Michelle Craig, Andrew Peterson, Sarah Peterson Following a thread: Knitting patterns and program tracing Craig, Peterson, and Peterson go on to say that programmers can learn from the way the online knitting community has developed standard conventions for communicating concepts like iteration, conditions, and documenting design decisions. Specifically, the way they've come about while trying to make the patterns understandable to beginners. For example, using termination cases with "until" makes the while loop easier to understand. They also say that creating a knitted final object is analogous to tracing, as it demonstrates an understanding of the pattern. Conclusion: Programming and Knitting are Very Similar Having taught myself programming and knitting and crochet, I can safely say the mindset required is very similar. From figuring out how to make a project work as you go to regular expressions in patterns, there are a lot of parallels. Programming as it is today would not exist without fiber, and programmers could learn a thing or two about teaching beginners from the online knitting community. I hope you enjoyed this brief overview of the history of fiber arts and programming. For general recognition and brevity's sake, I stuck to knitting, but my first fiber love will always be crochet. I personally believe in crafting as therapy for anyone, and we could all use a little self-care these days. If you've never thought of yourself as a fiber artist, but have a penchant for coding, try picking up some yarn and tools. You may be surprised. If you're already a programming fiber artist with Ravelry projects, check out my project, ! Knitworthy

Knitting Helps You Understand Programming Patterns 

While consumer products get all the hype, B2B products are a quiet gem “What’s cool nowadays?” I love asking my friends this question and hearing about the latest products and brands they’re loving. I’ve also noticed that in all my years of asking this question, I’ve never heard anyone mention an enterprise company. Can you imagine hearing a response like, “Oh — Cloudflare, Databricks, Superhuman, Slack, Salesforce, Envoy, the list goes on!”? To be fair, it’s easy to get lost in the consumer product bubble. Many of the most salient technology companies today are consumer internet companies. They’re splashy and (for better or for worse) attract all the hype. Consumer products also easier to identify with and more consistently used, so it makes sense that they occupy a prominent cultural position. But, after spending my career both building consumer and business products, it's become increasingly clear that building for enterprise is the easier and more pragmatic route. Here's why: It’s easier to tell when you have product market fit in enterprise. Consumers have much more varied needs than businesses. Consumers need to solve the nuisances of their daily life, but they also need to feel seen and heard, they need to be cool, they need to be entertained, they need to know about things before their friends do, they need to express themselves in a way that shows they’re woke but not overbearing or obnoxious, and on and on. Businesses, whose only real needs are to operate and make money for shareholders, are much less complex than humans. If they’re fiscally responsible, they will focus primarily on solving the problems that prevent them from achieving long-term survival. Consumers, on the other hand, have many “problems,” but in reality, most are non-vital. This leads to the lukewarm adoption of most consumer products. As a founder or product person, it’s easy to get stuck in the “ .” You have some usage, but people seem to feel “just ok” about your product. People aren’t lighting themselves up trying to spread the word about your thing. Many founders flounder at this stage and mosey on for years before throwing in the towel. It’s frustrating, but it’s also the result of solving a problem nobody cares enough about. trough of sorrow When building B2B products, it’s easy to know if you’ve attained product-market fit because most enterprise products require you to pay for them. If you’re not doing a good enough job at solving a need, companies won’t hesitate to churn, since churning actively costs money. This means that when enterprise customers use your product, they’re largely looking to answer two questions: not Is this product solving a problem of mine? Is the problem being solved so difficult or important that I would pay to not have to solve it myself? Enterprise problems are problems of survival, while most consumer problems aren’t. Since the former is born from desperation, while the latter is a nice-to-have, B2B customers are much more incentivized to find a solution that works. If yours doesn’t, they’ll churn. You then, as the product person, can get signal quickly about what you’re doing right and what’s going wrong. Consumer product usage is subject to factors like trendiness, taste, and psychology, creating inconsistency and volatility. Consumers are inherently fickle, which makes success hard to sustain. Even if you emerge from the trough of sorrow to find what feels like product-market-fit, your product can still fall to the wayside as new fads materialize. Take Yo, a great example of how consumer products can disappear just as quickly as they’ve risen. The app, which will forever live in infamy as a product whose only functionality was to let people say, “yo” to their contacts, had, at one point, almost 3 million registered users and over 1 million monthly active users. Yo, an app once valued at $10M. Source: The Verge It’s meteoric ascent was grounded in a novelty that eventually wore thin as people came to realize, “yo” was really not all that interesting. While Yo is a rather obvious example whose demise seems easy to predict in hindsight, others like Flappy Bird, Vine, Yik Yak and Peach represent products that simply lost their luster for no apparent reason. The risk of being upended by variable external factors is the price of building consumer products. Just imagine being that data scientist. B2B products don’t face the same volatility because they’re less swayed by cultural zeitgeist. You may stop using a product because none of your friends are using it or because it’s no longer trendy, but businesses remain unfazed in the face of these factors. They’re much more rational. There are clear moral exceptions to this that may cause companies to jump ship from even the best of products (e.g. CEO of an enterprise software company becomes embroiled in workplace harassment lawsuits), but otherwise, most enterprise customers will continue to pay so long as the solution they’re paying for continues to work. What’s more, corporate clients also tend to move slower than individuals. If a company decides not to renew your software offering, they may wait until the end of the quarter or pay period to mobilize their employees to stop using it, as opposed to deleting you from their phones the next day. This flattens the curve and gives you the runway to bolster your pipeline or improve your product to make up for lost business before it’s lost. You’ll never get a warning when an individual ditches your social app for the next TikTok. For consumers, too many risk factors are out of your control. It’s hard to predict what’s trendy and what has lasting power. And when you think you have it, you might win fast only to fail faster. Hitting on something that will stand the test of time requires a lot of luck. Enterprise execution is a straightforward path. One of the benefits of building paid products is that if people are unhappy with your product, they’ll find a way to let you know. When I worked on building Nearby Friends at Facebook, a free location feature enabling you to see which of your Facebook friends were around your area, we would conduct user research by incentivizing people to attend focus groups or take surveys with Amazon gift cards. We’d beg people to participate, since one of our biggest concerns was not getting a large enough sample size to inform our decision making. For people, attending these sessions was usually a big chore. You could tell most people didn’t really want to be there. When I transitioned to working on paid enterprise features, I saw the opposite phenomenon. People were clamoring to give us feedback because they were paying for our product and wanted to do anything in their power to get their money’s worth. We instrumented a feedback button on every page and, to my surprise, businesses would take the time to report bugs, write long, open-ended pieces of feedback, and request new functionality through the channel. Businesses have a greater sense for what they want. They’re able to identify gaps that, if solved, would better serve their use cases. If you asked a consumer, “would you want an app that lets you see photos posted by your friends, supports direct messaging, and shows you ephemeral photos and videos that expire after 24 hours?” (Instagram) would they say yes? Or if you had asked a consumer 10 years ago whether they would stand in line for hours to buy a phone with a 3X larger screen (iPhone), would there be consensus? Actual consumer behavior is much more inconsistent with what people report they want. This makes product execution circuitous and often unfruitful. It’s easy to build the thing people say they want and sit back and watch as no one uses it. Businesses report feedback based on real pain points. There is a world of growth opportunity that lies simply in solving these pain points and expanding to adjacent issues/use cases. And if you get things wrong, you can bet that companies will find a way to scream at you over the phone about taking away something that was beneficial or adding something that made their lives more complex. In this way, execution for B2B products is much more linear. In the enterprise world, there are also more consistent ways of operating and getting to success. When you don’t need to hack virality or devise creative marketing tactics for shock value, your work becomes less guesswork of what will capture consumer attention, and more about how to build an enduring and scalable product. Interestingly, most enterprise companies are structured the same way across sales, marketing, engineering, etc., so learnings are broadly applicable across B2B companies, even if the products being sold look markedly different. While the success of consumer products is hard to trace to a specific set of factors given the impact of trends and psychology, enterprise products have a relatively consistent playbook. Enterprise products make for better businesses. The challenge of building consumer products is not only that of creating a sticky user experience, but that of generating the monetary runway to continue iterating. While it’s true that some of the most lucrative and outsized exits have been consumer businesses, these should be considered outliers. Why? Let’s take a look at the most common monetization schemes: Advertising (e.g. Google, Facebook) In-product purchases (e.g. buying add-ons in games) Subscription (e.g. Netflix, Underclub) Commissions on selling product (e.g. selling physical goods or collecting app install fees for software) If you’re building a consumer product, monetization is difficult before you’ve achieved significant scale. Each of these models has their own associated challenges. You can’t build an a business based on advertising if you don’t already have millions of people using your product. If you offer in-product purchases, you’ll need to sell something that’s high-margin and re-buyable. Subscriptions are so common they’ve begun creating for consumers that don’t want to be beholden to so many monthly payments. And finally, selling products won’t create an outsized consumer outcome unless you are marketing an extremely high margin product (e.g. luxury goods). fatigue Even businesses like Youtube and Uber, which have grown to millions of daily users, still face tough questions on the viability of their business models. It’s not uncommon that SaaS companies like Salesforce and Shopify see gross margins of >85%, while salient consumer players like Netflix, Snapchat, and Uber generate gross margins between 40–55%. Enterprise products are usually subscription services, which businesses will pay for so long that it solves their problems. And unlike consumers, businesses don’t fall prey to subscription fatigue. We can also look at startup exit data to understand the ease of value creation across consumer and enterprise companies. Since 1995, the number of enterprise venture-backed exits has consistently outpaced consumer exits (4,600 enterprise vs 2,600 consumer), though in recent years, the value of individual consumer exits has eclipsed enterprise exits fairly consistently. Source: Sapphire Ventures What does this tell us? It’s much more common to see enterprise companies succeed. It’s hard to build consumer businesses, but if you are one of the lucky few to reach the promised land, the prize is certainly worth grinding for. Like most things, the decision of whether to build B2B or B2C comes down to your goals. If your goal is to learn how to develop strong products and businesses, enterprise is probably the right choice. On the other hand, if you’re excited by fame and can handle extreme frustration, maybe you’re cut out for the harsh climb of the consumer world. The best thing to do is to make a determination of what matters to you, stick to it with resolve, and guide yourself down the path most aligned with your inner scorecard.

Why Building Enterprise Products Is Awesome

re you looking for a way to create a lightweight real time analytics engine?This article explains how to do exactly that by combining a fast, embed-able real time analytics library (WSO2 Siddhi) with an immensely popular micro services platform, Spring Boot. As a bonus I will show you how doing that under 5 minutes using with code-generation!. After following this article you can generate a full-fledged Java spring boot project with a documented REST API, friendly, responsive front-end, basic spring security, comprehensive test coverage and database integration with just a few shell commands. The generated code will be your foundation for real time micro service/ application. A Jhipster is a handy application generator that creates Spring Boot and Angular application. Jhipster is one of the best low code Development platforms in open source community. Jhipster had become very popular GitHub within short amount of time. It has a high performance Java stack on the server side with spring boot and a front-end with angular,bootstrap and react. It makes project management easy with powerful workflow , build tools like Yeoman, Webpack and Maven/Gradle. Many developers personally use it to generate multiple Spring micro services that are preconfigured to work in their company’s infrastructure. But this articles’ objective is not to give a detail explanation about Jhipster. This article will focus on how to create a Jhipster module for WSO2 Siddhi. If you are new to Jhipster or want to read more about Jhipster, please read { }. https://cbornet.github.io/ You may have some specific real time analytics (WSO2 Siddhi) setup that you like to use in your Jhipster projects. Because you don’t want to reinvent your own wheel in every project, it makes sense to abstract all the boilerplate into your own generator. In that case you can build your own Jhipster generator for WSO2 Siddhi and use it with your every Jhipster project. At , we heavily rely on code generation (with a custom built tool chain) to get rid of a lot of boilerplate code which we tend to program into each and every solution/project/product. When we encounter a repeatable architectural pattern we tend to apply our tool chain so next time we don’t have to code it again. Most of this work, including the tool chain is part of the Xiges low code platform. Xiges.io Getting set up This section explains how to build the basic Jhipster generator module. If you already know how to create a jhipster module, you can skip this part and go to the next section. A Jhipster module is a Yeoman generator. Also, it is an NPM package. As a first step to generate a basic yeoman generator (which would eventually be our jhipster generator) install . After that you’ll need to have , and installed as well as the generator (yo) for creating generators.Please follow below commands with npm. NodeJS Yeoman Yarn Bower npm install -g generator-generator npm install -g yeoman npm install -g yo To check whether Yeoman is installed correctly just type command on the command line which would list you all the installed generators. If you can find yeoman there, you are good to go. Finally make sure you have git installed.First create a new directory which you’ll write your jhipster generator. This directory must be named as . Normally a Yeoman generator is prefixed with “ but Jhipster modules are prefixed with . Executing the below commands in order would create the folder and generate the jhipster module template. yo generator-jhipster-<name of your module(Siddhi) > generator-” “generator-jhipster-“ npm install -g generator-jhipster-modulemkdir generator-jhipster-siddhicd generator-jhipster-siddhiyo jhipster-module This is not really a JHipster module, that is meant to not be used in a JHipster application. This module is used to generate a new template for coding a new JHipster module for Siddhi. A JHipster module to create a JHipster module. Answer the questions to generate the module, so we can implement the changes. File structure As you can see we have readme file and package.json for generator itself. Test folder holds tests for the generator. The index.js file is the entry point for the generator. It contains the template files for the boilerplate (for generating the actual scaffolding). We created the default generator. Now we can modify it and add the our custom features. Scripting the WSO2 Siddhi generator Now we will check how to customize the jhipster generator we created and add our own features in the generator with the following steps: Setup and import the generator-jhipster The index.js file needs to export the generator-jhipster which will get run by Yeoman. Now I am going to clear the everything in our generator and start from the scratch. Here is what index.js file look after that. const chalk = require(‘chalk’);const generator = require(‘yeoman-generator’);const packagejs = require(‘../../package.json’); // Stores JHipster variablesconst jhipsterVar = { moduleName: ‘siddhi’ }; // Stores JHipster functionsconst jhipsterFunc = {}; module.exports = generator.extend({ // all your yeoman code here }); We assign the extended generator to the make it available to the ecosystem. This is the typical method we use when we export modules in the Node.js. Then we can have our own functionalities to perform through methods. In every method we added to the index.js is run once the generator is called and usually in sequence. Some method names have priority in this generator. The available priorities (in running order) are in order: module.exports initializing- Initialization of your methods.(getting configs and checking project current state ) prompting — prompt user preferences for the options( call this.prompt()). configuring — Saving configurations and configure the project. default — If your method name doesn’t match priority and put into this group. writing — copy template-files to the output folder and parsing (routes, controllers, etc). conflicts — Handling the conflicts. install — Where installations are run and specially add maven dependencies to the target Jhipster project pom file. (npm, bower) end — Called last, cleanup. After implementing these methods your file should be like this. Jhipster module uses , which is one of the main functions in Yeoman. The “composeWith” (this.composeWith()) method in yeoman generators allows the generator to run parallel with another generator and it can use features from the other generator instead of having to do it all by itself. In this above example it composes with the “jhipster:modules” sub generator and get the access to Jhipster’s variables and functions. Composability As you can see in the code we add Jhipster function (under the install() phase) to add Maven dependencies into the pom.xml file (addMavenDependency). Also used Jhipster global variables. These are the short definition for each variable. baseName: the name of the application packageName: the Java package name angularAppName: the AngularJS application name javaDir: the directory for the Java application, including the            package folder resourceDir: the directory containing the Java resources (always src/main/resources) webappDir: the directory containing the Web application (always src/main/webapp) You can see java spring boot classes and resource files, add to the ()phase and it’s done by calling the template function(all the java spring boot classes described in the next section ) . If you interested in other functions available in Jhipster module please visit — writing https://cbornet.github.io/modules/creating_a_module.html. 2. Initializing the generator We initialize our generator with package.json. In the above code Node invokes that require() function with a local file path as the function’s only argument and it gets the package.json file.This file is a Node.js module manifest. Our package.json file must contain the following. I configured the following: yo: CLI tool for running Yeoman generators. generator-mocha: a generator for the mocha test-framework. gulp: a front-end build tool. mocha: mocha is a JavaScript test framework for Node.js programs. 3. Create gulp file We use the gulp file as our build system. Tools like Gulp are often referred to as “build tools” because they are tools for running the tasks for building a web application.Your gulp file must be look like this. It’s often used to do front end tasks like: Spinning up a web server Reloading the browser automatically whenever a file is saved Using preprocessors like Sass or LESS Optimizing assets like CSS, JavaScript, and images So now we have finished the creating basics of jhipster generator for WSO2 Siddhi. But now we have to configure Spring boot with Real time analytics(WSO2 Siddhi). Spring boot + Real time analytics Spring boot is a very popular java-based framework for building web and enterprise based applications. Spring framework provides a wide variety of features addressing modern business needs. Spring boot makes it easy to create stand-alone, production-grade spring based applications that can “just run”. Spring boot takes an opinionated view of spring platform and combining third party libraries so developers can start with minimum fuss. Most of spring boot applications need a little bit of spring configuration. Because of this spring boot had become low code development platform. So It’s easy to embed WSO2 Siddhi to spring boot. Lets see how it’s done. Before we go further we have to understand main concepts about real time analytics with WSO2 Siddhi. WSO2 Siddhi is a java library which carries out real time processing on complex events.The streaming SQL Language of Siddhi is being used to describe complex conditions from the data streams. Siddhi is able to perform both Stream and complex event processing. Below diagram shows the basic workflow of the WSO2 Siddhi 3.0. Basic workflow of the WSO2 Siddhi 3.0 Now we are Embedding Siddhi in a Java Spring boot project and it allows you to use the Siddhi query language to carry out real time processing of complex events without running a WSO2 CEP server. Implementing Business Service and adding POST Rest Service for Siddhi Application As a first step we need to define a stream definition and Siddhi query.Stream definition always define the format of your incoming events and query is defined as below. String definition = "define stream TempStream(roomNo int, temperature double, deviceId long);" String query = "@info(name = 'avgTemperature') " +"from TempStream#window.time(60 sec)  " +"select avg(temperature) as temperature,deviceID " +"group by roomNo " +"insert into AvgTempStream ;"; Step 02: Creating Siddhi runtime This step involves creating a runtime representation of a siddhiAppRuntime by combining the stream definition and the Siddhi query you created in Step one. SiddhiManager siddhiManager = new SiddhiManager();//Generating runtimeSiddhiAppRuntime siddhiAppRuntime = siddhiManager              .createSiddhiAppRuntime(definition+query); The Siddhi Manager parses the Siddhi App and provides you with a Siddhi app run time. This Siddhi app run time is used to add callbacks and input handlers to the Siddhi app run time. Step 03: Registering a Callback Siddhi has two types of callbacks. Stream Callback — this subscribes to an event stream Query Callback — this subscribe to a query We need a callback to retrieve output events from query. So we can register a callback to the Siddhi app runtime. When results are generated, they are sent to the receive method of this callback. Also, we can print the incoming events from an event printer which is added inside this callback. ex-: siddhiAppRuntime.addCallback("AvgTempStream", new QueryCallback() {@Overridepublic void receive(Event[] events) {EventPrinter.print( );}}); events Step 04: Sending events As a final step you need to send events from the event stream to the query, you need to obtain an input handler as below example. //Retrieving input handler to push events into Siddhi InputHandler inputHandler =siddhiAppRuntime .getInputHandler("StockEventStream"); //Starting event processing siddhiAppRuntime.start(); //Sending events to Siddhi inputHandler.send(new Object[]{2, 23.0, 100L}); Refer to these files at the bottom of the article for exact implementation of the Business Service and adding POST Rest Service for Siddhi Application. Conclusion Creating a JHipster module with WSO2 Siddhi is an easy way to simplify your micro service generation, especially if your micro service uses the same configuration. Since it is a module, it’s very easy to add functionalities and meet your needs. Our company uses this real time analytics for making real-time predictions and tracking with our IOT products. Xiges Solutions Here is GitHub repository with the module used in this blog. Feel free to fork it and make changes to match your company requirements! Now you can download jhipster generator for siddhi by using this command npm install g generator-jhipster-siddhi Then run the module on a jhipster generated application. yo jhipster-siddhi Soon this module will be available in the jhipster marketplace.(After JHipster team verifies it ) link for the source code- https://github.com/xiges/generator-jhipster-siddhi. link for the npm registry- https://www.npmjs.com/package/generator-jhipster-siddhi. If you got any suggestions or enhancements you want to see in the module, please create an issue in the Xiges git hub repository — https://github.com/xiges/generator-jhipster-siddhi/issues That’s all I have for this topic. Thanks for reading. Until next time! References https://yeoman.io/authoring/composability.html https://www.jhipster.tech/modules/creating-a-module https://www.baeldung.com/jhipster https://yeoman.io/authoring/file-system.html https://gulpjs.com/docs/en/getting-started/javascript-and-gulpfiles https://siddhi-io.github.io/siddhi/#try-siddhi-with-wso2-stream-processor https://docs.wso2.com/display/CEP400/SiddhiQL+Guide+3.0

Lightweight real time analytics with Spring Boot + WSO2 Siddhi

Whiteboard coding interviews can cost you a job wow this article did have quite a few claps, getting closer to 17k! I’m grateful to everyone who showed support! Just to give you an update, I’m now a software engineer at ASOS and could not be happier than this. Well actually I’ll be even happier in a few days, when my first kid will be born 😍I have many ideas for new articles, I will just need to find the time to write them. Want to get notified? is the best way! march 2019: Adding me on Twitter Recently I’ve been quite close to getting a job at one of the . I went through the screening process and the smoothly but I failed to pass one of the final stages, a set of one to one, or two to one, interviews: Big 5 take-home assignment : algorithms. Whiteboard coding interview Technical interview: , CS, React. JavaScript Culture fit: that. : this maybe in another article :P Second whiteboard interview It went wrong There are a few mistakes that contributed to this temporary debacle. I may have passed the technical test and the culture fit (let’s just assume that) but I admittedly performed badly at the whiteboard coding interviews. Thinking in retrospective about that day I couldn’t expect any different outcome.The truth is: . I’m not prepared to do whiteboard interviews Whiteboard interviews ( ) The whiteboard interview is, in some ways, a hybrid of a technical test of the depth of your knowledge and also a social exercise source We all know what it is, right? it’s not much about the code, it’s about your solving abilities, it doesn’t matter if etc etc… I’m a frontend developer, I generally don’t implement interview-like algorithms on the fly or articulate my thought process while I code, I sort of make interactive UIs most the time with algorithms. When I need to implement one that I didn’t use for a while, I just start researching. How much do these tests tell about a candidate anyway? even said: Someone In real life, you would rarely just whip up an off-the-cuff algorithm from memory in the middle of a coding session. You are almost always going to use an existing library, which has its own test suite, and has survived the scrutiny of other developers. The only world where you would actually need to be able to recall an algorithm would be a post-apocalyptic one, where the hard drives of all the computers connected to the internet were fried, and all copies of foundational academic papers and computer science textbooks had been reduced to ashes. My opinion tend to match the one of the aforementioned author although I like to believe that such a skill (being good at whiteboard interviews) is about having a set of very good to have other skills, soft and hard. Mitigating circumstances first I’ll make this list because you may relate to some of them but most of all I still have to vent a little bit 😄 I didn’t do the interview in my first language. Under the spotlight, it gets harder. I’m a self taught frontend developer. I lack an academic preparation. I didn’t do many interviews in my career. And not many of them had whiteboard interviews kind of tests. I don’t do much public speaking. Unfortunately, at the moment, not much. You could object that none of these are actually mitigating circumstances and you would be right. By definition, mitigating circumstances are out of our control: the truth is that I could , , , . improve my English do more CS do more interviews do more public speaking I’m gonna try to solve that first whiteboard test the way I should have done that day and I’m going to try to report the process here. The Test ['Tokyo', 'London', 'Rome', 'Donlon', 'Kyoto', 'Paris']

// YOUR ALGORITHM

[    [ 'Tokyo', 'Kyoto' ],    [ 'London', 'Donlon' ],    [ 'Rome' ],    [ 'Paris' ]] That’s it. If you rotate the letters of each city you may or may not match another city. In case you do, put them together in a array on their own. At work you’re escaping labyrinths of thousands of lines of code on a daily basis, how difficult is to show these people how to solve this simple and even kinda funny problem? Hold my beer. Yeah… Let’s pretend it went slightly differently. The perfect whiteboard interview Catchy paragraph title, although wrong. This is just the way I could have solved this test, both code and thought articulation. Who know how many other ways there are! thank you very much. I’ll gladly solve this problem for you (little bow) Me: First question, do the letters just ‘rotate’ or they can be randomly mixed up? Me: (to protect the privacy of the interviewers, I changed their name into a fancy one): Just rotate. First letter goes last, etc. Larry One at a time. IMPORTANT NOTE There are a few comments of people proposing their solution to this problem. First of all thank you, one day I’m gonna look at them all and do something with them :) But it’s not the point of the article to get the most performant solution out of this. Also please pay attention to my last question to the interviewer. Many skip this part when they comment with their solution: letters rotate If you do so, combinations like and would match: they shouldn’t. Tokyo matches only with: okyoT, kyoTo, yoTok, oToky. See? 5 letters, 5 possible combinations. There’s an hint for you there. Let’s continue. one by one. It’s not enough to .split and .sort the letters. tokyo yookt right. Me: right. Larry: Right! I guess to start I would need a way to rotate the letters of each city. Tokyo would became okyoT, then kyoTo, oh cool! we’ve got kyoto now! Ok, I’m going to need a function to do that. Me: I also need a way to loop through the input cities, ‘rotate’ the letters, do some matching and lastly grouping them together. I see that the input and the output are both arrays and they both contains the same values, the only difference is the depth of the two arrays, a flat one and an array of arrays. Me: I’m gonna write some pseudocode now to validate my thoughts. Later I’m going to test it too, cause I’m that good. 😏 Still me: function groupCitiesByRotatedNames(cities) {

  // use a variable to contain the output.  let output = [];

  // loop through each of the cities  //  rotate the name in any possible combination  //  check if the output contains a city that matches a combination  //   add that city to the array containing the match  //   otherwise add the city to the output as a new array

  return output;} Me, confident as a rockstar: let’s pseudotest my pseudocode (😛). const input = ["Tokyo", "London", "Rome", "Donlon", "Kyoto", "Paris"];

console.log(groupCitiesByRotatedNames(input));

// That's how it would build up. The final output would be the last array:

// [//     ["Tokyo"]// ]

// [//     ['Tokyo'],//     ['London']// ]

// [//     ['Tokyo'],//     ['London'],//     ['Rome']// ]

// [//     ['Tokyo'],//     ['London', 'Donlon'],//     ['Rome']// ]

// [//     ['Tokyo','Kyoto'],//     ['London', 'Donlon'],//     ['Rome']// ]

// [//     ['Tokyo','Kyoto'],//     ['London', 'Donlon'],//     ['Rome'],//     ['Paris']// ] Let’s start implementing this algorithm. First, the utility to rotate the letters of the city names. I will create a utility function: Me: let’s make this function a bit smoother (and less readable). Overconfidence at its peak. Right! I like the reduce method, I’m gonna use it again! (It’s also very functional-programmy so it’s cool to use in an interview 😎) Me: The method applies a function against an accumulator and each element in the array (from left to right) to reduce it to a single value ( ). **reduce()** MDN Our single value being the desired output. Let’s sum up what this algorithm does: We check for each city in the cities array, we rotate the name and we match every combination with the accumulator value. If we find something we add the new city to the array containing the matches (with ) otherwise we just a new array containing our new city. Array.splice push Et voila’: All together now! . Get the Gist here but this is what came to my mind that day.(It’s complexity is quadratic while you may want to solve it with an algorithm that takes linear time O(n)). This solution is not the best one My actual whiteboard interview I started from the solution. I said, ok, I’m gonna need a reduce method. I immediately approached the board to write actual code without really knowing the plan. I did have an idea of the solution, pretty much the one I wrote in the previous chapter, but I didn’t it, I didn’t really it consistently to my poor interviewers. So I got lost in the code, I lost the the train of my thoughts several times, I also stumbled on syntax errors, in a mix of poor written pseudocode and actual code. There was no way to turn the tide at this point. A performance like this can make a lot of damage to the outcome of the interviews, especially if you let this affect the rest of the process on your side. articulate explain What should I do now Practice, practice, practice. Get a whiteboard, pick a problem and talk out-loud to the air and write, do that a lot. Practice the whiteboard interview, learn the routine. Your routine, any routine. Learn it like a song on the guitar, a card trick or whatever dangerous stunt with your skate. You need to prepare a speech, it’s like a presentation. It’s like a presentation with variables. A variable is the problem. I’ll go like this: I’ll spend some time analysing the requirements (1 min) I’ll formulate questions if necessary and get answers (3 min). I’ll pause and think of a direction to take (few min, don’t be afraid to stay silent for a while). I will propose initial solutions and get feedback from the interviewers (5 min). I will pause again and choose a solution (2 min). I will write pseudocode (5 min) I will test my pseudocode (5 min) I’ll finish by transforming the pseudocode into actual code. (until necessary) In around 30 min I will have solved the problem in a enjoyable and ordinated way. Conclusion PS There is also an interesting discussion around this article here: https://dev.to/albinotonnina/how-to-lose-a-it-job-in-10-minutes-35pi Resources ! wrote a wonderful comment and she allowed me to paste here a list of resources that she put together following some whiteboard interviews. I found this list very very useful! Update Evana Mp She says: After this experience, I made a list of resources which I collected from many web Q&As related to this and in order of priority this is the list of books, sites, and courses I will be working with: Books: Cracking the coding interview Introduction to Algorithms Programming Interviews Exposed Clean code Algorithm Design Manual How to Think About Algorithms Programming Pearls Data Structures And Algorithms Made Easy In Java 2nd Edition 2nd Edition Websites: InterviewBit HackerRank Codewars Codefights AlgoExpert CarrerCup Interview Cake LeetCode Geeks for geeks Pramp HackerEarth HiredInTech Courses: Coderust 2.0 Thanks for reading And Thanks and for proof-reading this :) Adam Mackintosh Josh Comeau Hello, my name is Albino Tonnina, I’m a frontend engineer who does bad whiteboard interviews at the moment and works in London, you can find me on or or or around the city. Twitter Github Instagram My latest articles: Designers are from Saturn, developers are from Jupiter: or, why communication matters Speaking of web layouts…introducing the Magic Hat technique 🎩✨

How to Lose an IT Job in 10 Minutes

Random title, I know, but aren't you just a little bit intrigued? I love a teaser title that makes you go ‘hmm’ and then *fingers crossed* sucks you in as you thirst to know more. To elaborate, ‘the talk at MOO’ happened earlier this week on Tuesday and was a talk by James Kirkham, Co-founder of Holler and Global Head of Mobile and Social for Leo Burnett. His talk was titled ‘Social Media will never be the same again’ and was hosted by MOO, the company who print rather fancy cards and are based in Shoreditch #obvs business I hot-footed it to that talk with Digital Mums partner-in-crime, Annika, departing from the convivial Digital Mums shipping container office (yes, really) in Hackney. Despite the fact we were already in East London, we still managed to be a tad late #awks as we walked along Scrutton Street, firstly, missing the building completely (doh!) and, secondly, on entering, not realising the talk was in the building basement. The room was full when we arrived and James had already begun; I entered efficient mode, spied two stools in a meeting room off the kitchen, beckoned Annika over slyly so that the other two ladies we walked in with and didn't know wouldn't clock them and then we plonked them down by the ping pong table (naturally) and adopted that air of having been there t_he whole time._ James had a lot to say about social media and I took as many notes as I could in my trusty notepad with a purple-inked pen. Since a few of my colleagues couldn't make it, I felt it was my duty to soak up as much info as I could so I could share it later which is, well, what I'm planning to do now so here goes: ‘Effectiveness is now reliant on great ideas that people talk about’ James expanded on this further by citing the example of a Mercedes TV advert he had worked on that was a ‘socially driven advert’, appropriately hash-tagged with #youdrive. People were able to choose the ending of the advert by tweeting with the hashtag and were also sent back a personalised film poster image with a Mercedes featured within it. He highlighted the increase of reactive content and warned that brands should not saturate social media and simply fill it with noise. Another interesting social media campaign that James talked us through was created for the Jura Whisky Distillery Company. The idea was created in an hour and cost in the region of £100 or so but the return was over £200,000 — impressive stuff, so what did they do?! In response to Google Maps somehow taking the Isle of Jura, where the distillery is based, off the map, they launched #findjura. The buzz around ‘the missing island’ was covered in the media and people seem to enjoy playing the devised ‘pin-the-tale-on-the-donkey’ style #findjura game. I must say, I preferred this example to the Mercedes one as it was for a lesser known brand, didn't have a massive budget behind it but using the power of a catchy hashtag and bags of personality, it was still able to have a huge impact. Isn't that what the power of social media is all about? James is, unsurprisingly, a massive enthusiast and, honestly, the fervour of his speech on its rapid growth meant I was hardly able to keep up with my note-taking at times! One of the key points that stuck with me, however, is that when technology is completely adopted, the word initially used to describe it disappears as it becomes so embedded in our lives. To illustrate this, he said that in ten years time, the word ‘smartphone’ wouldn't be used any more and I must say I'm inclined to agree. technology He gave us some rather impressive internet use stats to ponder: 14 years ago, 400m people use the internet Today, 3bn people use the internet with 2bn accessing the internet via their smartphone thus demonstrating the fusion of the internet with mobile telephony and why he stated earlier in the talk that It sure is. . ‘your social strategy is your mobile strategy’ This neatly segwayed into social media and apps we use on our phone. He cited the seemingly comic but ultimately highly successful example of ‘Yo’, an app that allows you, with the push of a button to send the word ‘yo’ to your contacts. The app was created by one person with zero investment and has gone on to have a million users. The point James was making about ‘Yo’ was that with its one-touch-one-function facility, the app played into what we are increasingly seeking — simplicity. He then stated: with the backdrop of an image of a Swiss Army Knife for good measure. ‘The app that does everything is dead’ No talk about social media and mobile would be complete without a little nod to Snapchat, a social media platform that is way out of my demographic though I feel more and more curious about trying it out. Snapchat is all about being in the now, James said, while adding that it’s great for sneak peeks and teasers and is fun and exciting. Testino even shared some photos of Burberry’s new collection on Snapchat. He spoke to some teens about their interaction with Snapchat and it seems they find it authentic due to the increased honesty shared as a result of its transient nature. He also told us that young people used ‘Snapchat Discover’ as a means of seeking advice as they were drawn to its visual nature. On this note, James told us about Tunepics, an app that essentially allows users to share pictures/photos with music over the top and the inclusion of a mood wheel. In case you hadn't noticed, emotion on social media is a big deal and it’s probably going to continue that way for some time and that’s not to mention the ever increasing use of emoji in our everyday interactions these days. I'm usually partially to the cheeky wink one most of the time but can’t say I've ever used the aubergine in earnest (we all know what it stands for right?!) James touched on Periscope and highlighted the unpredictable success of #DrummondPuddleWatch (people watching other people navigate and/or jump into a massive puddle in Drummond!) as well as mentioning the relatively recent addition of Moments on Twitter. He also added that with WhatsApp, the rise and rise of messaging will present opportunities for brands to think about especially as these messages become increasingly audio/visual. His speech continued at 100mph as he discussed ‘haptic experiences’ through the use of technology you wear, affective interfaces that allow you to judge the mood of the nation and the body as an interface through the use of wearable technology. I must say that the ‘Toxic Friend Finder’ bracelet did sound pretty out there but, hey, you never know… As he summarised the key learnings to take away, he reminded us that our social media content needs to be: Bespoke Personal Tailored Relevant As well as the fact that: ‘social is everywhere, it permeates everything’ Rounding off with: ‘Your ideas inspire technologists. There’s never been a better time to be in the ideas business’ It’s no wonder that once this talk was done, I found myself grabbing handful after handful of the free popcorn around and refuelling after taking all that in. As people started to leave, Annika and I found our Digital Mum mates and a trip to the pub was arranged to debrief. We couldn't possibly divulge what we talked about there because, hey, when mums get a night out, you the conversation gets very interesting indeed. What I will tell you, however, was that a cat with a diamante collar came into the pub, strode along the bar and plonked itself on a barstool and naturally I took a photo and tweeted it as #standard. The internet still loves a cat pic, after all. know

Digital Mum Days: The talk at MOO

Computer vision remains an extremely compelling application of artificial intelligence. Whether it’s recognizing elements on a battlefield or predicting crop yields, computer vision is arguably one of the most commercially valuable (and societally important) realms of AI. However, a rate-limiter to the adoption of the best computer vision capabilities is oftentimes the complexities associated with constructing a dataset and designing a simple end-to-end system that will perform your computer vision task on a new image. In this blog post, we will take a step-by-step look at how to tackle these issues with best-in-class tools like CVAT and MinIO Bucket Notifications. By the end of this post, you will be able to train an object detection model on a custom dataset and use it to make predictions whenever a new image appears. The Premise Let’s say we want to be able to recognize types of aircraft present in satellite imagery. Let’s also assume we are starting from scratch: no pre-built datasets, no pre-trained models. Here are two sample aircraft that we want to detect and recognize in our satellite imagery: The steps outlined in this post can be generalized to just about any domain. Instead of detecting aircraft types, we could be classifying land use or performing regression to predict crop yields. Beyond traditional images, we could also be training and performing inference on other types of multi-dimensional data like LiDAR point clouds or 3D Seismic Images; it becomes just a question of how the training data looks (and potentially a different deep learning model instead of YOLO). If you have more questions about what this would look like for a specific use case, feel free to make an issue on the GitHub ! repo Step 1: Acquiring and Managing Training Samples For this project, due in large part to not owning an on-demand imaging satellite, I visited airfields on Google Earth and took multiple screenshots of areas that had some of these planes visible. Assembling this set of images took quite some time, so I stored them all in a bucket on my MinIO server titled “object-detection.” In a production setting, the benefits of storing your collected samples on MinIO become even more prescient. Active-active replication, the highest levels of encryption, and super fast GET/PUTs (to name a few) mean that your diligently-collected samples will be highly available, safe and secure. Step 2: Creating the Dataset In order to train an object detection model for your use case, a labeled (or ‘annotated’) dataset is needed. A great tool for this is CVAT by OpenCV. A cool feature is that CVAT provides a utility to connect your MinIO bucket as a “cloud storage” in order to feed your bucket’s images directly to the dataset annotation tool. To do this, make sure that your MinIO Server’s host is accessible to the CVAT server, especially if you’re running MinIO Server on-premise or locally on your laptop. Also, as a note, there are two ways to use CVAT: (1) using their provided web app at or (2) running it locally. In either case, once you have CVAT open, click on “Cloud Storages” in the menu bar. From there, you can fill out a form to attach your (S3-Compatible) MinIO bucket: app.cvat.ai Let’s now create our new labeling task under “Tasks”: You should be prompted with a form to fill out: When creating the task, it’s important to define the class labels correctly (I defined two Rectangle labels titled “SU30” and “TU95,” corresponding to the two planes I wanted to detect): Now the remaining step is to attach our previously added MinIO bucket as a data source. Under “Select Files,” click “Cloud Storage” and fill in the name you provided for that source earlier. I used the name “minio-cv-bucket” above. The uploading process will take a few minutes. Once it’s completed, you should be able to see your annotation job available under “Jobs.” Now, by clicking on the job, you can start annotating each of your images. Warning: this can be a disproportionately time-consuming process. Generally, in a production environment with large annotation needs, it may be best to offload this task to a dedicated internal team or a third-party data labeling company. Once you’re done annotating, export the dataset in the YOLO format. Step 3: Organizing the Training Data Your exported dataset will be in the form of a zip file. Once you unzip it, the YOLO-formatted annotation text files will be within an enclosed folder. Feel free to take a look at them. In the YOLO format, each image’s annotations are in a text file where each line contains two corners of a bounding box and the class. The class number corresponds to the order in which you defined the labels when creating the task. So, in this example, 0 would correspond to Su-30, and 1 would correspond to Tu-95. At this point, create a new working directory (or enter one that you have already created). Within this directory, create a subdirectory called 'dataset.' Within ‘dataset’, create directories such that your working directory looks like this: my_cv_project (WORKING DIRECTORY)
|---- dataset
	|----images
		|----train
		|----val
		|----test
	|----annotations

		|----train
		|----val
		|----test You will now have to populate the train, val, and test subdirectories for both images and their corresponding annotations (the text files). It’s up to you how you want to retrieve and split up your samples. A good practice is to break up your total amount of training samples into 80% training, 10% validation, and 10% test. Make sure to randomly shuffle your images before partitioning them. Personally, I used MinIO Client’s in the command line to quickly retrieve all the images from my ‘object-detection’ bucket. Alternatively, if you have all your sample images in one place on your local computer already, you can directly work with that. Once I had all my samples in one place, I used a Python script to shuffle, split, and move my images and annotations to the train, val, and test directories. . If you have any questions about how to use it, feel free to make an issue in the ! mc cp Here is the script provided for convenience repo Ultimately, make sure that for each image that you have placed in either images/train, images/val, or images/test, the matching annotation .txt file is also in the corresponding subdirectory within the annotations/ directory. For example: my_cv_project (WORKING DIRECTORY)
|---- dataset
	|----images
		|----train
			- 5.png
			- 3.png
			- 2.png
		|----val
			- 4.png
		|----test
			- 1.png
	|----annotations
		|----train
			- 5.txt
			- 3.txt
			- 2.txt
		|----val
			- 4.txt
		|----test
			- 1.txt Now, our data is in place. It’s time to take a look at our object detection model and start training. Step 4: The Object Detection Model The current gold standard (in terms of performance and ease of use) for object recognition is the YOLO (You Only Look Once) class of models. At the time of writing, YOLOv8 is the latest version and is maintained as open-source by Ultralytics. YOLOv8 provides a simple API that we can leverage to train the model on our newly created annotations (and eventually run inference as well). Let’s download YOLOv8: $ pip install ultralytics We can now use the YOLOv8 CLI tool or Python SDK to train, validate, and predict. Refer to the YOLOv8 docs for further information. Step 5: Training In your working directory, define a YAML file that specifies the locations of the dataset and the details about the classes. Notice how the paths are the same as the ones I created earlier in the working directory. I called my file ‘ .’ Also, note the two aircraft class labels have to be defined in the same order as they were in CVAT. objdetect.yaml train: ./dataset/images/train/ 
val:  ./dataset/images/val/
test: ./dataset/images/test/

# number of classes
nc: 2

# class names
names: ["SU-30","TU-95"] Start training the YOLOv8 model on our dataset with the following command (using the YOLO CLI tool). Refer to the YOLO docs to learn more about all the different options you can configure for training. Here, I am initiating training for 100 epochs and setting an image size of 640 pixels (all our training images will be scaled accordingly during training): $ yolo task=detect \
  mode=train \
  model=yolov8s.pt \
  data=objdetect.yaml \
  epochs=100 \
  imgsz=640 Training will take a while, especially if you’re working on a laptop (like I am), so now is a good time to take a break (or read ahead 😀)! At the end of the training loop, your trained model, along with other interesting graphs and charts, will be stored in an auto-generated directory called ‘runs.’ The terminal output (like below) will indicate the specific location of the results of the latest run. Every time you train a model, a similar directory will be generated within ‘runs/detect/’. Results saved to runs/detect/train Note: runs/detect/train/weights/ will contain the PT files with the exact trained weights. Remember this location for later. Step 5B: Model Validation and Test You can run validation with the following command: $ yolo task=detect \                                       
  mode=val \
  model=path/to/best.pt \
  data=objdetect.yaml The results will be automatically stored in a folder in your working directory with a path of the form ‘runs/detect/val.’ To perform inference on the test set, you can use the following command: $ yolo task=detect \                                                    
  mode=predict \
  model=path/to/best.pt \
  conf=0.5 \
  source=dataset/images/test The results will be stored in ‘runs/detect/predict.’ Here are some prediction results on the test set: Step 6: New Image Inference using MinIO Bucket Notifications Now that we have a trained model that can recognize some aircraft types present in a satellite image, ? how can we make use of it for new images in a simple way is a perfect tool for this. We can build a system that can automatically perform object detection inference upon a new image dropped in our bucket with the help of a webhook. MinIO Bucket Notifications At a high level, we have 3 steps. First, we need to define an endpoint that can serve as a webhook to perform object detection on a new image with our trained model. Second, we need to configure some environment variables for our MinIO Server deployment that instruct it to hit our webhook endpoint upon some event happening. Third, we need to configure what types of bucket events (i.e. PUT) we want to act on. Let's walk through it step-by-step. Here’s the code for a simple Flask-based server ( ) that runs inference on a new image added to the MinIO bucket: detection_server.py """
This is a simple Flask inference server implementation that serves as a webhook for the event of a new image being added to a MinIO bucket. Object detection using YOLO will be performed on that image and the resulting predictions will be returned.
"""
from flask import Flask, request, abort, make_response
from ultralytics import YOLO
import tempfile
from minio import Minio

# Make sure the following are populated with your MinIO details
# (Best practice is to use environment variables!)
MINIO_ENDPOINT = ''
MINIO_ACCESS_KEY = ''
MINIO_SECRET_KEY = ''

model = YOLO('/PATH/TO/best.pt')  # load a custom model (path to trained weights)

client = Minio(
    MINIO_ENDPOINT,
    access_key=MINIO_ACCESS_KEY,
    secret_key=MINIO_SECRET_KEY,
)

app = Flask(__name__)

@app.route('/', methods=['POST'])
async def inference_bucket_webhook():
    """
    This endpoint will be called when a new object is placed in your inference bucket
    """
    if request.method == 'POST':
        # Get the request event from the 'POST' call
        event = request.json
        bucket = event['Records'][0]['s3']['bucket']['name']
        obj_name = event['Records'][0]['s3']['object']['key']
        with tempfile.TemporaryDirectory() as temp_dir:
            temp_file_name = temp_dir+'/'+obj_name
            client.fget_object(bucket, obj_name, temp_file_name)
            # See https://docs.ultralytics.com/modes/predict/ for more information about YOLO inference options
            results = model.predict(source=temp_file_name, conf=0.5, stream=False)
            # A list of bounding boxes (if any) is returned.
            # Each bounding box is in the format [x1, y1, x2, y2, probability, class].
            result = {"results": results[0].boxes.data.tolist()}
            print(result)
            resp = make_response(result, 200)
            return resp
    else:
        abort(400)

if __name__ == '__main__':
    app.run() Let’s start the inference server: $ python detection_server.py
 * Serving Flask app 'detection_server'
 * Debug mode: off
 * Running on http://127.0.0.1:5000
Press CTRL+C to quit Make note of the hostname and port that the Flask application is running on. Next, let’s start working on configuring the webhooks on the MinIO side. First, set the following environment variables. Replace <YOURFUNCTIONNAME> with a function name of your choosing. For simplicity, I went with Also, make sure that the endpoint environment variable is set to the correct host and port for your inference server. In this case, http://localhost:5000 is where our Flask application is running. ‘inference.’ $ export MINIO_NOTIFY_WEBHOOK_ENABLE_<YOURFUNCTIONNAME>=on 
$ export MINIO_NOTIFY_WEBHOOK_ENDPOINT_<YOURFUNCTIONNAME>=http://localhost:5000 Now, restart the MinIO Server using the command or if you’re starting the server for the first time you can also just use the command. For more information about re/starting the MinIO Server, check out the MinIO docs. Note: ALIAS should be replaced with the alias for your MinIO server deployment. For more information about how to set an alias or view existing aliases, check out . mc admin service restart ALIAS minio server the docs Finally, let’s add the bucket and event we want to be notified about. In our case, we want to be notified about ` events (creation of new objects) in our bucket. I made a brand new, empty bucket for this purpose titled so I will substitute that for ‘BUCKET.’ put` “detect-inference,” $ mc event add ALIAS/BUCKET arn:minio:sqs::<YOURFUNCTIONNAME>:webhook --event put You can check that you have configured the correct event type for the bucket notifications by verifying whether " " is outputted when you run this command: s3:ObjectCreated:* $ mc event ls local/detect-inference arn:minio:sqs::<YOURFUNCTIONNAME>:webhook For a more detailed explanation of publishing bucket events to a webhook, check out the . We are now ready to try object detection on a brand new image! docs Trying out our Inference System Here’s the new image (titled ‘1.png’) I want to conduct inference on: I drop the new image in my ‘detect-inference’ bucket: Almost instantly, I’m able to see the following results on my Flask server: $ python detection_server.py
 * Serving Flask app 'detection_server'
 * Debug mode: off
 * Running on http://127.0.0.1:5000
Press CTRL+C to quit

image 1/1 /var/folders/xf/q7x0z8sn5nvckccp1g0m1vpm0000gn/T/tmpo6jx3w8u/1.png: 448x736 2 SU-30s, 101.0ms
Speed: 4.1ms preprocess, 101.0ms inference, 5.8ms postprocess per image at shape (1, 3, 448, 736)
{'results': [[1927.78369140625, 627.7123413085938, 1995.090576171875, 715.3443603515625, 0.8142037987709045, 0.0], [1735.740234375, 477.2108154296875, 1809.181640625, 555.767578125, 0.7766116261482239, 0.0]]}
127.0.0.1 - - [14/Sep/2023 15:39:21] "POST / HTTP/1.1" 200 - Note that each detected bounding box in the results list is in the YOLO format [x1, y1, x2, y2, probability, class]. Here are the bounding boxes and predicted classes superimposed on the original image: Note: For production environments and/or large machine learning models, it’s a good idea to use an established model serving framework like PyTorch Serve or Triton Server to make inference more robust and reliable. If you’re interested in that, check out the previous post on . Optimizing AI Model Serving with MinIO and PyTorch Serve Concluding Thoughts We did it! We walked through how MinIO and CVAT came together to keep our collected image samples safe and available, as well as how to create our custom object detection dataset. Then, we trained our own custom YOLO model for our custom task. Lastly, with just over 50 lines of code, we put together an inference server using MinIO Bucket Notifications that could run a new image past our custom trained object detection model. Furthermore, for most mission critical applications of computer vision, it is best to conduct inference on the edge. Otherwise, said applications are vulnerable to the latency associated with uploading new data to the public cloud and waiting for an inference server in the cloud to get back with an answer– not to mention the risks of a faulty network connection. For this reason, a computer vision pipeline centered around MinIO as the data layer makes much more sense. Imagine a drone flying over an airfield, being able to capture, store, and use our trained model on new imagery with completely-on-board hardware and software. With a local deployment of MinIO server, the Bucket Notification-based inference system we built at the end of the post works perfectly for this scenario and countless others like it. If you have any questions join our or drop us a note at . We are here to help you. Slack Channel hello@min.io Also published . here

How to Train an Object Detection Model Using a Custom Dataset With MinIO and YOLO

MinIO এবং YOLO এর সাথে একটি কাস্টম ডেটাসেট ব্যবহার করে কীভাবে একটি অবজেক্ট সনাক্তকরণ মডেলকে প্রশিক্ষণ দেওয়া যায়

So trainieren Sie ein Objekterkennungsmodell mithilfe eines benutzerdefinierten Datensatzes mit MinIO und YOLO

Cómo entrenar un modelo de detección de objetos utilizando un conjunto de datos personalizado con MinIO y YOLO

Comment former un modèle de détection d'objets à l'aide d'un ensemble de données personnalisé avec MinIO et YOLO

मिनिओ और योलो के साथ कस्टम डेटासेट का उपयोग करके ऑब्जेक्ट डिटेक्शन मॉडल को कैसे प्रशिक्षित करें

MinIO と YOLO でカスタム データセットを使用して物体検出モデルをトレーニングする方法

MinIO 및 YOLO와 함께 사용자 정의 데이터 세트를 사용하여 객체 감지 모델을 훈련하는 방법

Como treinar um modelo de detecção de objetos usando um conjunto de dados personalizado com MinIO e YOLO

Как обучить модель обнаружения объектов с использованием пользовательского набора данных с помощью MinIO и YOLO

MinIO ve YOLO ile Özel Bir Veri Kümesi Kullanılarak Nesne Algılama Modeli Nasıl Eğitilir

Cách đào tạo mô hình phát hiện đối tượng bằng cách sử dụng bộ dữ liệu tùy chỉnh với MinIO và YOLO

如何使用 MinIO 和 YOLO 的自定义数据集训练对象检测模型

This application integrates YOLOv3 and PolyRNN++ for comprehensive image analysis. The App and Demo: Try the App!! https://objsegment-a5090a8bb17f.herokuapp.com/ https://www.youtube.com/watch?v=ZneiGaACYN4&embedable=true Why Image annotations? Image annotation is the process of labeling or adding metadata to images to provide additional information about the contents of the images. It involves the process of labeling images with relevant information, such as object boundaries, semantic segments, or object classes, to train these models effectively. It plays a crucial role in enabling machine learning models to understand and interpret visual data. However, the task of image annotation is not only laborious but also time-consuming, often requiring significant human effort to annotate each image accurately. Annotating a single object within an image can take up to 40-60 seconds, highlighting the challenges and costs associated with this process. The market for data collection and labeling is growing rapidly, with a projected value of . $2.82 billion by 2023 To address these challenges, the objective of this project is to automate the semantic segmentation process as much as possible, aiming to minimize the manual labeling effort required. By implementing a annotation tool, the aim is to develop a system that can accurately detect and segment objects within images automatically. Such an automated system would not only reduce the burden on human annotators but also improve efficiency and potentially reduce errors in the annotation process. deep learning-based Object Detection Object detection entails the identification and localization of objects of interest within an image or video. Its primary objective is not only to classify the objects present in the image but also to determine their exact spatial coordinates, generally in the form of bounding boxes. There are various model architectures that have been developed for object detection, each with its own strengths and limitations. A few of the commonly used architectures: RCNN — Paper & Code Fast-RCNN — Paper Faster-RCNN — Paper YOLO — Paper (You only look once) is a single-shot object detection algorithm, i.e., a single neural network that predicts bounding boxes and class probabilities directly from full images in one evaluation. This unified model enables faster predictions, and unlike sliding window techniques, YOLO considers the entire image and simplicity encodes contextual information about the classes and their appearance YOLO . YOLO models object detection as a regression problem and works as follows: Divides the image into an SxS grid. For each grid cell, target variable y is: a) B bounding boxes represented by the center of the bounding box relative to the grid cell ( x, y coordinates), width (w), and height of the box (h) with respect to the whole image. b) Confidence for those boxes, which reflects how confident the model is that the box contains an object and how accurately the box has predicted the object boundaries. Confidence = Pr(Object)*IOU. c) C class probabilities: C class probabilities P (Class I | Object). The overall target (y) for the image becomes an SxS(5B + C) tensor. Train a convolution network with input as an image and output with a tensor of the dimensions y. The output of the final layers then goes through that selects a single bounding box out of many overlapping bounding boxes. Non Maximum Suppression a) Discard low probability predictions. b) Compute the IOU of boxes; for boxes with IOU ≥ 0.5, retain the one with the highest probability. In summary, YOLO’s innovative approach to object detection has made it a go-to choice for real-time applications. Its ability to handle multiple objects in a single pass, coupled with its high accuracy and speed, has made it an indispensable tool in the field of computer vision. In recent years, YOLO has continued to evolve, with versions like YOLOv2, YOLOv3, and YOLOv4 addressing various challenges and improving object detection performance even further as the field of computer vision advances, YOLO and similar algorithms are likely to play an increasingly critical role in shaping the future of AI-driven applications. Semantic Segmentation Semantic segmentation divides the image into distinct regions based on the semantic meaning of the objects within it. When it comes to representing the segmentation, there are two primary options: a pixel-by-pixel approach, where each pixel is classified as foreground or background, or a sparse polygon representation that outlines the object boundaries. There are various model architectures that have been developed for semantic segmentation, each with its own strengths and limitations. Few of the commonly used architectures: U-NET — Paper & Code Mask-RCNN — Paper Polygon-RNN — Paper and its variants comprise two key modules. The first module focuses on the task of image representation and the extraction of image features. In this module, a Convolutional Neural Network (CNN) is employed, specifically utilizing the VGG network architecture without a dense layer. To enhance the information flow, skip connections are incorporated at the top of the network, enabling the fusion of information originating from different levels of the CNN. This fusion is crucial as it combines both low-level details, which are essential for identifying edges and boundaries, and high-level features, which play a pivotal role in recognizing objects within the image. This processed image representation is then passed into a Recurrent Neural Network (RNN), which employs Convolutional Long Short-Term Memory (ConvLSTM) units. The role of this RNN is to generate a sequence of vertices that collectively define the object of interest in the image, effectively providing a structured and accurate representation of the object’s shape. Polygon RNN Polygon RNN’s ability to generate polygonal representations offers a valuable advantage, especially when dealing with complex object shapes and fine-grained segmentation tasks. By using sequences of vertices to outline object boundaries, it strikes a balance between computational efficiency and segmentation accuracy. In conclusion, semantic segmentation plays a pivotal role in computer vision applications by providing a detailed understanding of image content. Model architectures like Polygon RNN and its variants showcase the innovation in this field, enabling the accurate delineation of object boundaries. As research in semantic segmentation continues to advance, we can expect more sophisticated techniques and models that further enhance our ability to extract meaningful information from images, opening up new possibilities for AI-driven applications across various domains. References: Yolo v3 Architecture — https://towardsdatascience.com/yolo-v3-explained-ff5b850390f Redmon, Joseph, et al. "You only look once: Unified, real-time object detection." . 2016. Proceedings of the IEEE conference on computer vision and pattern recognition Poly-RNN++ Architecure — https://arxiv.org/pdf/1704.05548.pdf ADE20K Dataset — https://groups.csail.mit.edu/vision/datasets/ADE20K/ Also published . here

Revolutionizing Image Analysis: YOLOv3 and PolyRNN++ Integration for Image Annotation

An Experiment coauthoring with ChatGPT. Where Ana and Olivia chat about Yatoshi and Hal Finney. When cryptography meets quantum computer, and cryopreserved bodies hide secrets worth millions it only takes one bad actor to unleash a thrilling action. Coauthored with ChatGPT based on my prompts. The technology and historical events in this novel are real. Previous chapters can be found here. Chat between Ana and Olivia : Ana, I’d like to introduce you to Olivia. She’s also very interested in quantum computing. Peter : Great to meet you, Olivia. Ana : Nice to meet you too, Ana. Peter has told me a lot about your work in the field. Olivia : Really? That’s kind of him. What about you, Olivia? How do you earn your bacon? Ana : I focus on bitcoin mining and cryptography in general Olivia : That’s really interesting. I’ve been working on using quantum computers to rescue lost bitcoin keys. Ana : Ana, I heard that you were able to recover the keys that Peter lost to his wallet. How did you do it? Olivia : Ever heard of the “If I tell you I might have to kill you“ line?  But seriously, I used a combination of classical algorithms and quantum computing techniques to recover the keys. First, I generated a large number of candidate keys using a classical algorithm, then I used a quantum computer to quickly eliminate the invalid keys. Ana : That’s impressive. How did you ensure that the quantum computer was accurately checking the validity of the keys? Olivia : I used Grover’s algorithm, which is specifically designed for searching unsorted databases. It allowed me to quickly find the correct key among the candidates, even though the database was very large. Ana : That’s amazing. I’ve heard of Grover’s algorithm, but I haven’t had the chance to apply it in practice yet. Olivia : It’s a very powerful tool. And with the advances in quantum computing, it’s becoming more and more accessible for real-world applications. Ana : Absolutely. I can see it being very useful for cryptography and code breaking. Olivia : Olivia, that sunset in your latest picture is breathtaking. Where was it taken? Ana : Thank you! That was taken by the sea in Bali. I’m actually here for the Bitcoin conference. Olivia : Wow, that’s amazing! Bali is such a beautiful place. How is the conference going so far? Ana : It’s been fantastic! There have been so many interesting talks on the latest developments in blockchain technology and cryptocurrency. And the location is just incredible. I feel so lucky to be here. Olivia : That sounds like an amazing experience. I’ve always wanted to attend a conference in an exotic location like Bali. Ana : You should definitely try to make it happen sometime. It’s not just about the conference, it’s also about exploring new places and meeting new people. Olivia : Well, you do look great in that pic. You kind of remind me of my ex-girlfriend. Ana : Oh really? Well, If I may say it, you are looking good yourself too. Olivia blushes before answering There is a small awkward silence in the chat. Olivia starts typing again. : You know, Peter told me a story about someone named Yatoshi who made a big impact on the world of cryptocurrency. Have you heard of him? Olivia : Yes, I have. Yatoshi is widely considered to be the pseudonym for the person or group of people who created Bitcoin. Ana : That’s right. It’s amazing how something that started as a simple idea has grown into a global phenomenon. Olivia : Definitely. And speaking of the early days of Bitcoin, have you heard of Hal Finney? Ana : No, I haven’t. Who is he? Olivia : Hal Finney was an early adopter and developer of Bitcoin. He’s most famous for being the recipient of the first-ever Bitcoin transaction. He received 10 Bitcoins from Yatoshi himself. Ana : So what happened to Hal Finney? Did he continue to work on Bitcoin and other projects in the industry? Olivia : Unfortunately, Hal passed away in 2014. But his legacy lives on. He was a true pioneer in the world of cryptocurrency, and his contributions to the development of Bitcoin have had a lasting impact. Ana : That’s sad to hear. It’s always difficult to lose someone who has made such a big impact. Olivia : Yes, it is. But his memory will always be with us. And in a way, a part of him will always be here. He was cryopreserved after his death, so there’s a chance that he may be able to be revived in the future. Ana : That’s fascinating. I’ve heard of cryopreservation, but I didn’t realize that it was being used in cases like this. Olivia : Hal died from a disease called Amyotrophic Lateral Sclerosis, or ALS. It’s a degenerative illness that affects the nerve cells responsible for controlling muscle movement. Ana : That’s terrible. So he was unable to continue his work on Bitcoin and other projects in the industry. Olivia : Unfortunately, yes. But despite his illness, his brain was unaffected, and that’s why his family decided to have him cryopreserved. They hope that one day in the future, he may be able to be revived and cured. Ana : That’s a beautiful thought. To think that he may be able to see the world he helped create, and all the ways in which his ideas have changed the world. Olivia : Cryopreservation is a growing field, and there are a lot of people who believe that it has the potential to change the way we think about death and what it means to be alive. But regardless of what the future holds, Hal Finney will always be remembered as a true visionary and a pioneer in the world of cryptocurrency. Ana : It’s been great chatting with you, Ana. I’ve learned so much about the world of cryptocurrency and the amazing people who have helped shape it. Olivia : The pleasure has been all mine, Olivia. You have such a great passion for this industry, and I think that’s what sets you apart. Ana : Thank you, Ana. You’re too kind. Olivia : So, since we’re both in the field of technology, I was wondering if you’d like to come visit my lab after the conference. I’d love to show you the latest advancements in quantum computing and see what you think. Ana : That sounds amazing! I’d love to see what you’re working on and learn more about your research. Olivia : Great! I’ll make sure to send you all the details once I get back to the lab. And in the meantime, enjoy the rest of the conference. I can’t wait to hear all about it when you get back! Ana : Thanks, Ana. It was really nice chatting with you and I can’t wait to meet you! Olivia Also published here.

Stories, News & Mentions about YOTO

Yoto’s Stories on HackerNoon

Yoto's News & Mentions