Deep Sentinel is a Pleasanton-based emerging startup building a unique next-generation technology solution for home security

Deep Sentinel

How new technologies are redefining its boundaries Since the beginning of rapid human evolution and expansion, our security as an individual and as a species has been the key driver for our landmark inventions. The was to protect ourselves from wild beasts and germs from raw food. The was to secure our journeys so that we could travel longer distances with less fear and more adventurous spirit. The came out of a primal need for security in terms of permanence of location. We can quote many more examples, but it is clear that security of self and/or community is what motivates us for achieving progress. discovery of fire and its control invention of the wheel necessity for shelter In this day and age, however, security has taken a more nuanced and subtler meaning to all of us. Sure, we still install expensive security measures in our homes to protect our valuables and property (unless, of course, you’re Kevin McCallister from the series). Home Alone But, nowadays, security includes emotional and psychological dimensions, as well as legal and ethical standards. Eminent people like M.K.Gandhi and Martin Luther King Jr. have fought for the freedom and security of their respective oppressed classes. Even now, we have many parades to defend and equalize the LGBTQ community around the world. In light of this, we are also witnessing the emergence of new technologies with rapid scope for expansion like Artificial Intelligence, Machine Learning, Blockchain etc. Also, we are witnessing the destructive effects of technology intrusion into our lives, such as with the Facebook data leaks or the Cambridge-Analytica scandal. In this scenario, we must look at how these technologies can help or harm our notions of security, how they could possibly broaden our very definition of security, and a few personal ideas of mine about how they could be of huge benefit to us in the long run (or even in the short run). Fair warning though, I am NOT an expert at any of these technologies. I’m just a normal guy, like you, trying to make sense of a rapidly changing world, using my own limited knowledge to navigate through this labyrinth. Artificial Intelligence & Security Photo by on Franck V. Unsplash How It Could Redefine Security Nowadays, artificial intelligence is progressing rapidly, whether it be in increased interactions with the user based on past experiences (cognitive intelligence), ability to understand what the user is feeling based on stimuli cues (emotional intelligence), or in grasping the external environment to tailor the response it gives even further (social intelligence). However, how far can this go? As AI systems continue to reach human standards in complexity, it’ll start to develop a new form of security i.e. security for programs. And I’m not even talking about firewalls and antivirus, it’ll soon involve security guidelines for robots as well. A basic example is the fundamental Three Laws of Robotics devised by the famous science fiction writer Isaac Asimov and made popular by the dystopian representation in “I, Robot”. However, there have been many papers and surveys written on the topic of ‘Robotic Ethics’. For instance, a conducted in the year 2011 by Mr. Pawel Lichocki and his team from the Learning Algorithms and Systems Laboratory ( ) of talks about the main problem in ethics (when a robot causes harm) being figuring out who is responsible when such an incident happens, the humans operating it or the robots themselves. More recently, there has been an International Conference on Robotic Ethics and Standards (ICRES) 2018 where the keynote speeches have been about creating a modern standard in AI systems, about autonomous weapons and the future of war etc. survey LASA EPFL Therefore, we can see that Artificial Intelligence can possibly branch out the definition of security to include human-robot and robot-robot interactions, which must be done as a safeguard against dystopian futures and unfavourable utilisation by antisocial third parties. How It Is Currently Helping Achieve Better Security AI is finding a wide variety of applications in terms of personal security as well as detecting potential threats. The main application for AI in this sphere is in identification of possible threats (which have been already uploaded into its database) and then scanning the external environment (such as in malls, parks and other public places) to identify, analyse and pinpoint the suspect, thereby acting as a sort of preemptive strike or counter-attack in case of suspicious activity. To this effect, some technologies have already been developed to implement this. The first technology is , developed by , which is using AI to scan and profile all the persons passing through, say, a particular security checkpoint, using accurate visual sensors and face recognition technology. Once that is done, for each individual, the AI program cross-verifies it with its database for suspects and criminals, and currently, it can identify potential threats at the rate of 1–2 persons per second (which is quite a speedy process!) Evolv Edge Evolv Technologies The second prominent technology is offering a complete home security package which scans real-time for suspicious activity, recognizes it and alerts the police in addition to alerting the home owner via a customized mobile app. You can check about this security system . Deep Sentinel here Even in military purposes, drones equipped with AI are able to scan an environment for threats and are able to traverse terrains difficult for humans to reach as well, thereby increasing the reach of armies for surveillance, reconnaissance and counter-terrorism. A prime example is , a quadcopter-type drone developed by which is powered by Hivemind, a machine learning application which I will talk more about in the section on Machine Learning. Hivemind Nova Shield AI How I Believe It Can Help Security Photo by on Owen Beard Unsplash One of the key challenges when it comes to AI is defining the upper and lower limits i.e. what is the basic functions an AI should be able to do (which depends on the application it is to be used) and upto what extent should AI assimilate and interpret data accordingly (which depends on the amount of control a user wants to exert on it). This should be the first question to be answered, in my opinion, as valid answers specific to the application will help mitigate risks and errors due to either software or operation method. Another critical issue that could be addressed is the privacy levels that can be defined. In other words, we should know upto what extent can the AI predict or observe information, especially when dealing with sensitive issues involving people. For example, even in the home security system, we should be sure that the data the security system receives when it identifies ‘friends’ or ‘relatives’ should be stored isn’t shared in any external server, so as to prevent misuse by anybody. Machine Learning & Security Photo by on Franck V. Unsplash How It Could Redefine Security Nowadays, Machine Learning is taken in sync with Artificial Intelligence when it comes to real life operation or discussions regarding security. And rightly so, as AI algorithms need huge data sets to effectively respond in the field, which is facilitated by Machine Learning and Big Data. However, Machine Learning has already transformed how we view security as a concept, with the advent of the data leaks and the Cambridge-Analytica scandal. This happened partly because it tapped into our understanding and interpretation of security. In simpler terms, it made us realise that we wouldn’t be comfortable when some random program knows more about us than we do and we will feel insecure when said program uses said data to effectively change strategic decisions in a particular way. intuitive This expansion of understanding is also critically important when creating algorithms for medical purposes, as a person’s medical records are supposed to be confidential between the patient and the doctor. This involves thorough analysis into what goes in the program and also defining limits to the extent of invading patient’s medical data security for the purpose of more accuracy. In a broader sense, Machine Learning has now made us understand that there is a concept of ‘personal space’ even on the Internet domain, and that we should be mindful of our activity in that domain. Personally, I am afraid as to how my data is being used by some third-party program, but from a neutral perspective, it would be interesting (to say the least) to see how this redefines how we view Machine Learning as a concept, and indeed, how we could/would tweak programs to suit our purposes. How It Is Currently Helping Achieve Better Security Photo by on Luca Bravo Unsplash Most of the Machine Learning applications in terms of Security are done in sync with AI for better performance of the algorithm. AI algorithms are done with a combination of supervised learning (classification) and unsupervised learning (clustering) for collection and interpretation of its data set. However, the classification set is more predominant when starting of with an AI algorithm in its initial stages. When it comes to classification, the supervised part, there are 4 key phases:- A phase, where a model is constructed by applying a classifier to a set of training data training A phase, where the analyst applies the validated data to the model to assess the accuracy validation A phase, where this model is now tested using data that is new and NOT a part of the previous two phases, and finally testing A phase, where the sufficiently tested model is now used on fresh unaltered data. deployment In these phases, Machine Learning helps immensely in the training and deployment phases for the express purpose of collecting data, hence improving the accuracy of the model created, thereby shortening the duration of the validation and testing phases, leading to a shorter execution period overall. Also, like I had mentioned before, Machine Learning is also being used in military drone applications, with the example I had taken being . Hivemind In this case, Hivemind uses visual sensors to accurately map the environments and feed it as virtual data to the algorithm which in turn stores the data and uses it to understand different environments so that the next time it is sent to survey an unknown terrain, it scans and analyzes based on the previous information it has collected and learned from. How I Believe It Can Help Security Although Machine Learning is already used extensively, and in conjunction with AI as well, I think it can go further. There is a possibility it already exists, but we could use Big Data and Machine Learning to predict the next terrorist strike, or the next worldwide health epidemic. However, I think the best way that Machine Learning can help us is in forensic profiling. Law enforcement is one of the few areas where complete data can be obtained without fear of persecution, hence we could condense hundreds of thousands of archives regarding data on forensic evidence into a single algorithm that can track and provide clues that will help us stay two or even three steps ahead of the perpetrators of crimes. As of now, the most important part is resolving the issue of privacy of personal data in other fields like medicine and public safety, because that is a crucial viewpoint that needs to be addressed. Also, once done so, laws must be enacted to enforce strict rules for privacy to the public and to corporates alike. Internet-of-Things (IoT) and Security Photo by on Bence ▲ Boros Unsplash How It Could Redefine Security IoT has had its own share of problems regarding security, with some recent leaks having put serious doubts in the minds of consumers regarding the efficacy of IoT devices in remaining secure. Although there have been of Alexa from the Amazon Echo laughing randomly (?), one such IoT-based attack happened quite recently . unspecified reports when an unsecured product demo of LocationSmart , a geolocation data firm, allowed the user to look up the location of any mobile phone without providing any security credentials . This was especially dangerous as all the major mobile carriers in the US (like Verizon, AT&T, Sprint, T-Mobile etc.) were affected by this hack, including a couple of Canadian carriers as well So, IoT can potentially break down security if in the hands of the wrong people, good to know. But, is prevention and safety enough? We need to look into the basic concept of allowing toasters and vacuum cleaners to have microchips inside which can connect to the Internet and can be controlled with the touch of a button. Although I agree that we will have our lives easier and at the touch of a button, we should be concerned whether our button is the only one that can control our devices, because tech going out of control could potentially lead to household accidents as well, which is even more alarming . considering that household accidents result in the deaths of more than 18,000 people every year , that too in America alone Sure, we don’t need to be this pessimistic, but think about your own home appliances being used to do covert surveillance, which can be used for blackmail or worse. Hence, we need to be very careful, as IoT, if misused, has the potential to create huge (yet hidden) security breaches, whose damage can be irreparable. How It Is Currently Helping Achieve Better Security Photo by on Andres Urena Unsplash Although I have mentioned about and how it is touted to be an effective security system for your home, IoT as a means of security is still in its nascent stages, and the situation is such that IoT requires its own security instead of it providing the same. Deep Sentinel There are many methods being followed for safety of IoT devices, with the simplest being IoT authentication by means of a simple password, or a two-step verification, or biometric identification etc. Also, methods like IoT encryption and IoT network security are being implemented, but then again, . And it will be a long time before IoT can act as a viable standalone security measure. we are creating security for IoT, instead of it securing us How I Believe It Can Help Security To be honest, I think is going in the right direction regarding this, but my idea is that if we can use AI to monitor and defend against cyber threats for IoT devices, it would be a better countermeasure than a simple password or an antivirus. Moreover, instead of connecting all devices to the cloud, we could have a local intranet for each home, where the requisite devices are connected to this intranet and then connected to the phone, and can only be accessed via a separate authentication system, again AI being the most preferred security option. When this occurs, IoT could be used safely and also effectively to provide home security efficiently. Deep Sentinel A second idea is that IoT for military weaponry could be implemented (or maybe it already is) but again, much care must be taken with AI to safeguard against external cyber threats, so that wars aren’t started suddenly. A third, more outlandish idea of mine is , like for example using AI to isolate a particular malware or virus and rerouting it to a particular device (here by device I mean a household appliance like a toaster) and then disposing of the toaster. I’m sure it’s not as easy as I make it to sound like, but it would be very helpful if we could find the source node of any cyber-attack in addition to isolating and storing that node inside even an IoT toothbrush, and then throwing the toothbrush away, so that it won’t have any point of re-entry. ( ) using IoT to store malware in a physical form and then disposing it physically Don’t quote me on this, its pure speculation! Conclusion A peek into the future. Photo by on Samuel Zeller Unsplash These new technologies are here to stay, and even as we speak, they are increasingly being utilized in our daily lives, to make us more comfortable, or at the very least, more secure. We do have a long way to go, and as the times change, our definition of ‘security’ is sure to undergo some major upheavals. However, we hope that with the right track of progress, and with the right set of laws passed and research done, we could be setting foot into a safer, more secure, yet free world. . And that is a future I’m keen to experience 👏 💓 Did you like this article? If so, please do (you can do this more than just once if you really enjoyed it) to show your and also consider responding to this article down in the comments section below if you feel like sharing your thoughts on this!! Thanks for reading! :) Also, do consider reading my other recent articles here . And, you can always contact me on Facebook and know about my professional life on LinkedIn . Have a great day!! :)

Security In This Day & Age

Hate Speech Detection in Algerian Dialect Using Deep Learning: Our Methodology

Th-th-th-that's right, folks! We've got ourselves a real doozy here! Two of the most titillating, tantalizing, and downright terrific terms in the tech world - AI and cryptocurrencies - sharing the same sentence! It's like a tech enthusiast's dream come true, I say, it's a dream come true! But hold your horses, this ain't just a figment of your imagination, no siree! It's as real as the snout on my face! And let me tell you, it's as exciting as a barn dance on a Saturday night! So, strap yourselves in, folks, 'cause we're about to embark on a rip-roaring ride through the thrilling intersection of artificial intelligence and digital currencies. It's a wild world out there, full of possibilities as vast as a cornfield in July. So, let's dive in, shall we? After all, as they say in the world of high finance (and I should know, being a pig and all), you've got to bring home the bacon! Okay, let’s get back to ‘serious mode’ - Just had to get this out of my system. The Role of AI in Cryptocurrencies Prepare for a deep dive into the intricate world of AI and cryptocurrencies. At first glance, they might seem like an unlikely duo. However, they're more intertwined than a pair of headphones left in your pocket. Artificial Intelligence, our digital Einstein, excels at deciphering complex patterns and crunching colossal amounts of data. In the volatile landscape of cryptocurrencies, this ability is as valuable as a diamond in the rough. Picture a vast ocean of data - numbers, trends, patterns, all swirling in a complex choreography. To the human eye, it's a whirlwind of chaos. But for AI, it's a playground. With its sophisticated algorithms and rapid processing power, AI sifts through this data with the precision of a master watchmaker, identifying trends and making informed decisions with the speed of a cheetah on the chase. And it doesn't stop there. AI also serves as a vigilant sentinel, tirelessly working to detect and deter cryptocurrency fraud. It's akin to having a highly intelligent, ultra-efficient assistant who never takes a coffee break, always keeping a watchful eye on your valuable digital assets. In essence, AI and cryptocurrencies, though an unconventional pair, are together reshaping the way we trade, invest, and navigate the digital economy. This is just the tip of the iceberg. As we delve deeper, you'll discover how the future of AI and cryptocurrencies is set to revolutionize the digital world. So, let's forge ahead! The Advantages of Integrating AI in Cryptocurrencies Now, let's cut to the chase. What's the real deal? What's in it for you? Well, integrating AI in cryptocurrencies has several advantages that are as clear as a bell: Increased Efficiency AI can automate trading and transactions, making them faster and more efficient. It's like upgrading from a bicycle to a rocket ship! For instance, a platform called leverages AI to create a hedge fund that allows thousands of data scientists to openly collaborate and compete to create the best predictive models. It's a clear example of how AI can increase efficiency in cryptocurrency trading. Numerai Enhanced Security AI can detect unusual activity and potential security threats, keeping your precious digital coins safe and sound. A study published on discusses how AI can enhance cryptocurrency security using a risk management model. The model uses social media indicators to mine threats based on investors' perceptions, thereby enhancing the security of cryptocurrency. IEEE Xplore Improved Accuracy AI can analyze market trends and make accurate predictions, helping you make informed investment decisions. It's like having a crystal ball, but way cooler! A platform called uses AI to predict Bitcoin prices. It's a prime example of how AI can improve accuracy in cryptocurrency predictions. BitcoinPredict In a nutshell, the integration of AI in cryptocurrencies is like adding a turbocharger to your car's engine. It boosts efficiency, ramps up security, and enhances accuracy. So, whether you're a seasoned crypto trader or a newbie, AI can give you a leg up in the crypto game. Now, isn't that something to write home about? The Challenges at the Intersection of AI and Cryptocurrencies Now, don't get me wrong, it's not all a bed of roses. Like any trailblazing technology, the crossroads of AI and cryptocurrencies has its fair share of speed bumps: Data Privacy AI is like a data vacuum, sucking up information faster than a kid with a milkshake. But who owns this data? And how is it used? These are big questions that need answering. For instance, the Cambridge Analytica scandal [ ] threw a spotlight on the potential misuse of data, raising concerns about how AI might use our personal information in the future. 1 Regulatory Issues Cryptocurrencies are still a bit of a wild west, and throwing AI into the mix is like adding dynamite to a bonfire. As KPMG points out, regulatory activity around crypto and digital assets is intensifying as usage by investors, companies, and even some central banks, shows widespread interest and adoption at retail and institutional levels Technological Hurdles Developing and implementing AI algorithms is no walk in the park. It's more like a trek up Mount Everest. It requires significant resources and expertise. For example, the concept of AI-complete problems, which are the most difficult problems in artificial intelligence and computation, poses a significant challenge in the integration of AI and cryptocurrencies Case Studies: Successful Integration of AI and Cryptocurrencies Alright, enough with the theory, let's get down to the nitty-gritty. Here are a few examples of projects that have successfully integrated AI and cryptocurrencies: Numerai This innovative platform uses AI to automate cryptocurrency trading, resulting in increased efficiency and reduced risk SingularityNET This groundbreaking initiative uses AI to detect and prevent cryptocurrency fraud, enhancing security for users Coinbase This pioneering venture uses AI to analyze market trends and make accurate predictions, helping users make informed investment decisions The Future of AI and Cryptocurrencies So, where do we go from here? Well, the sky's the limit! As AI and cryptocurrencies continue to evolve, we can expect to see even more exciting developments at their intersection. From smarter trading algorithms to more secure transactions, the future looks bright for this dynamic duo. Quick and Dirty: Alright, folks, let's cut to the chase. Here are the answers to some of the most burning questions about AI and cryptocurrencies: Q: Can AI predict cryptocurrency prices? A: Well, AI isn't a crystal ball, but it's the next best thing. It can analyze market trends and make informed predictions. But remember, cryptocurrencies are as unpredictable as a bull in a china shop. So, while AI can certainly help to make more informed predictions based on historical data and market trends, it's not foolproof. Q: How does AI enhance the security of cryptocurrencies? A: Think of AI as a digital watchdog, always on the lookout for unusual activity and potential security threats. It can help to prevent fraud and other forms of cybercrime, keeping your digital assets safe and sound. Q: What are some examples of projects that have successfully integrated AI and cryptocurrencies? A: There are several projects that have hit the bullseye. For instance, Numerai uses AI to automate cryptocurrency trading, and SingularityNET uses AI to detect and prevent cryptocurrency fraud. These are just a couple of examples of how AI and cryptocurrencies are joining forces to revolutionize the digital world. Jin Park Signing Off Well, folks, it looks like we've reached the end of the line. We've journeyed through the exhilarating crossroads of AI and cryptocurrencies, delved into the challenges and opportunities, and even dared to gaze into the crystal ball of the future. It's clear as day that the fusion of these two trailblazing technologies is set to shake up the digital world like a snow globe. But let's not forget, this is just the opening act. As AI and cryptocurrencies continue to evolve, so too will their intersection. So keep your peepers peeled and your noggin nimble, because the future of AI and cryptocurrencies is sure to be a rollercoaster ride that would make even the most seasoned thrill-seeker's head spin! We've taken a deep dive into the captivating world where AI rubs shoulders with crypto. It's been a hoot, and I hope you've enjoyed the ride as much as I have. We've laughed, we've learned, and most importantly, we've taken a leap into the future. So, as we part ways, remember to keep exploring, keep learning, and keep pushing the boundaries of what's possible in the digital world. After all, in the world of tech, the only constant is change. This is Jin Park, your guide in the digital wilderness, signing off. Until next time, stay curious, stay bold, and stay tuned for more adventures at the intersection of technology and imagination! Join the 'Jinius' movement with Jin Park. Dive into the digital realm now! Support the adventure at ❤️. Your contribution is the fuel for our journey!" ko-fi.com/jiniusme

When AI Meets Crypto: A Deep Dive into the Confluence of Machine Learning and Digital Currencies

Astounding Stories of Super-Science August 1931: VOL. VII, NO. 2 - The Danger from the Deep

Idea / inspiration I’m creating a project that can spot trees, using satellite data. I got this idea from the company . Which uses satellite imagery and AI to track forests and vegetation around the world. They are many uses for this technology. You can track deforestation in an area. By tracking the amount of missing trees over a certain timescale. A company called uses satellite imagery and AI. To prevent fires and power outages by spotting trees growing too close to power lines. 20tree.ai Spacept Using satellite imagery saves time and money. As inspecting forestry using a helicopter or drone can cost a lot of money. And will cost a lot of time, to set up the helicopter and drone to fly over the forestry. Dataset When I was first trying to get the dataset together for the machine learning model.  I settled on using . I got this idea from this . Which used satellite imagery to spot landslides. In the blog post he highlighted the issue of getting satellite data. Where if you wanted to get images from the . You will have to do them manually. And you cannot add coordinates of an area you want to capture. google earth engine medium blog post ESA Copernicus SciHub website So, you get continent sized images. Which I can agree from my personal use of the service as well. When I tried to get an image that covered a small area in my city. Instead I got large images covering all of France and Britain. With Google earth engine you can filter by date, area, and cloud cover. Which is why Google earth engine was recommended.  I used the code from the blog to create a dataset. I added customisations to the code for my dataset. The blog post described the process below: How our final (semi) automated pipeline ended up working was as follows: 1. Enter coordinates intocropping_coordinates.py to calculate the coordinates of a 10 km square. 2.  Copy the output and replace the variables at the top of the JavaScript code in the Google Earth Engine console (the JavaScript code we used). Don’t forget to change the date range to one that you’d like. 3.  Download the images with less than 100 for the mean cloud density given by clicking the link printed in the console output. When I got the images, they looked like this: Image by author In the beginning I did not know why this was the case. But later learned that satellite images were separated into different bands. Bands are certain wavelengths that the satellite captures. Doing this allows the user to view the image with features that are not visible with human eye or can be highlighted with a chosen wavelength. For example, a common use of satellite imagery is tracking farmland. By using the infrared spectrum, it is more easier spot vegetation. Figure 1 This false-colour image of Florida combines shortwave infrared, near infrared, and green light. (NASA by Matt Radcliff with data from the USGS ) image Landsat 5 Earth Explorer. Also found here: https://earthobservatory.nasa.gov/features/FalseColor gives a nice list of wavelengths and their uses: Wikipedia The wavelengths are approximate; exact values depend on the particular satellite's instruments: Blue, 450–515..520 nm, is used for atmosphere and deep water imaging, and can reach depths up to 150 feet (50 m) in clear water. Green, 515..520–590..600 nm, is used for imaging vegetation and deep water structures, up to 90 feet (30 m) in clear water. Red,600..630–680..690 nm, is used for imaging man-made objects, in water up to 30 feet (9 m) deep, soil, and vegetation. Near infrared (NIR), 750–900 nm, is used primarily for imaging vegetation. Mid-infrared (MIR), 1550–1750 nm, is used for imaging vegetation, soil moisture content, and some forest fires . Far-infrared (FIR), 2080–2350 nm, is used for imaging soil, moisture, geological features, silicates, clays, and fires. Thermal infrared , 10400-12500 nm, uses emitted instead of reflected radiation to image geological structures, thermal differences in water currents, fires, and for night studies. Radar and related technologies are useful for mapping terrain and for detecting various objects. Also, : combinations True-color uses only red, green, and blue channels, mapped to their respective colors. As a plain color photograph, it is good for analyzing man-made objects, and is easy to understand for beginner analysts. Green-red-infrared, where the blue channel is replaced with near infrared, is used for vegetation, which is highly reflective in near IR; it then shows as blue. This combination is often used to detect vegetation and camouflage. Blue-NIR-MIR, where the blue channel uses visible blue, green uses NIR (so vegetation stays green), and MIR is shown as red. Such images allow the water depth, vegetation coverage, soil moisture content, and the presence of fires to be seen, all in a single image. I wanted a simple RGB image. Which is simple visible light. As its best to start simple. To develop that, we need an image that combines all 3 bands into one. So, using some of the google earth engine example code and my personal changes. I was able to get some images. The red square covering the map is the area that google earth engine will capture for the image. Image by author The image below: Image by author As we can the image quality is horrible to be frank. I started to notice problems when setting up the red square. You see in the image of the general area of outside the red square. You can see that image lacks clarity. Which means if you zoom in even further you cannot tell what’s in the image because the quality is pretty bad. The reason for the problem was the resolution of images. The satellites sentinel-2 and Landsat only have a resolution 30 meters. This explains spatial resolution perfectly: website Spatial resolution refers to the size of one pixel on the ground. A pixel is that smallest ‘dot’ that makes up an optical satellite image and basically determines how detailed a picture is. Landsat data, for example, has a 30m resolution, meaning each pixel stands for a 30m x 30m area on the ground. It’s considered a medium-resolution image, which can cover an entire city area alone, but the level of detail isn’t fine enough to distinguish individual objects like houses or cars. Custom Dataset All open source data goes down to 30 meters. For finer data one must pay. So I had to find a solution that allows me get higher quality images. Without forking small fortune. After a lot thinking (and worrying). I found the solution hidden in plain sight. When using navigation systems on your phone. Your phone can give you a satellite image of your area. You can see which buildings are which. And differentiate between objects. So I decided to use (normal) google earth and capture images straight from the website. To make sure the screen captures where all the same size used a program called PicPick. Here’s one image I captured: Image by author Labelling data Now we have a small sample size. We need a solution for the machine learning model to know it ‘s looking at a tree. So used a program called Which I used to manually label the trees in the images. LabelImg from GitHub. Image by author I did not label every single tree in every image. As that will take forever. But wanted to have some labelling so the model can pick one or two trees in the test image. Same reasoning why the dataset is very small. The reasoning is so if I need to make changes to the dataset. It won’t take a long time. Also, just to create a model as simple prototype before scaling up the work. Model This project is an object detection problem. As I’m dealing with image data. I will be using some type of Convolutional Neural Network. One example I was thinking of using for the model is . But quickly learned that as the model was already trained the model won’t be able to read satellite image for trees. The blog post mentioned you do transfer learning or fine-tuning. But said t that it was outside the scope of the article. So, I had to find a new model to use. here R-CNN After a while of googling around. I found an interesting model that I can use called R-CNN. More precisely faster R CNN. R CNN which stands for Region-Based Convolutional Neural Network. Where the model selectively goes through many 2000 proposal regions of the image. After the proposal regions are selected, they then shaped for the input of the Convolutional Neural Network. After Support vector machines classify the regions. Then a bounding box regression is used to work out where the box goes. Figure 2 https://www.analyticsvidhya.com/blog/2018/10/a-step-by-step-introduction-to-the-basic-object-detection-algorithms-part-1/ But they are a few problems relating to using a traditional R CNN. Mainly that processing an image takes a lot of time. As the model is extracting 2000 regions to check. Prediction with the model takes around 40 seconds. Fast R-CNN A better version of R CNN was created by the same person. Called Fast R CNN. Main difference is instead of extracting thousands of regions from the image. Fast R-CNN inputs the whole image. Then the CNN produces a convolutional feature map from that the chosen region proposals. The region proposals are then shaped for the pooling layer. After that it gets reshaped again to be 4entered into a full connected layer. For each region of interest feature vector, it is moved to a SoftMax layer to predict the class. The use of bounding box regression to find the boxes. One problem is that as Fast R-CNN uses selective search is slowing down testing. But Fast R-CNN is way faster than the traditional R-CNN. As Fast R-CNN takes around 2 seconds to predict. Faster R-CNN Now we have faster R-CNN. The examples above uses selective search to go through proposal regions. Faster R-CNN uses a separate network of predict the region proposals. Using the convolutional feature map. Then the proposals are then reshaped for a pooling layer. For classification of the region. Also bounding box regressor used to place the images. Figure 3 https://www.analyticsvidhya.com/blog/2018/10/a-step-by-step-introduction-to-the-basic-object-detection-algorithms-part-1/ Coding up the of model I followed tutorial from . The tutorial used a dataset of blood cells. I replaced the dataset with my custom dataset of google earth photos. To get the data ready for the model I had to convert the bounding boxes from XML to CSV. here To do that I used pandas and python’s ElementTree module. And borrowed code from and the export file in the tutorial. machine learning mastery linked list_dataframe = []
folder = file os.listdir(folder):
    path = file = path + file
    print(file) filename = re.sub( , , file)
    filename = filename.replace( , )
    filename = filename.replace( , )
    tree_type = tree = ElementTree.parse(file) root = tree.getroot() boxes = list() box root.findall( ):
        xmin = int(box.find( ).text)
        ymin = int(box.find( ).text)
        xmax = int(box.find( ).text)
        ymax = int(box.find( ).text) row_list = [filename, tree_type, xmin, xmax, ymin, ymax]
        print(row_list) list_dataframe.append(row_list)


print(list_dataframe)

df = pd.DataFrame(list_dataframe, columns=[ , , , , , ])

print(df.head())

df.to_csv( , index= ) 'Custom_dataset\\Labels' for in 'Custom_dataset\Labels\\' # filename = os.path.splitext(file)[0]+'.png' 'Custom_dataset\\\Labels' "" "\\" "" ".xml" ".png" 'Tree' # get the root of the document # extract each bounding box for in './/bndbox' 'xmin' 'ymin' 'xmax' 'ymax' # coors = [xmin, xmax, ymin, ymax] # boxes.append(coors) # df = df.reset_index(drop=True) 'filename' 'tree_type' 'xmin' 'xmax' 'ymin' 'ymax' 'testing.csv' False df = pd.DataFrame( , columns=[ , , , , , ])
list_dataframe = []
folder = None 'filename' 'tree_type' 'xmin' 'ymax' 'ymix' 'ymax' 'Custom_dataset\\Labels' This section sets up the variables to be used for the loop. An empty dataframe is set up with the name of the filenames, class and the bounding boxes. The path of folder containing the images are saved in the folder variable. file os.listdir(folder):
  path = file = path + file
  print(file) for in 'Custom_dataset\Labels\\' The start of the for loop starts by listing the files in the folder. Then the file path then printed by combining the folder path and name of the file. This is done as a normal for loop without os.listdir did not work on my laptop. filename = re.sub( , , file)
filename = filename.replace( , )
filename = filename.replace( , ) 'Custom_dataset\\\Labels' "" "\\" "" ".xml" ".png" When printing the filename it had too many backslashes. So to get rid of it str.replace functions were used. Also, as the folder it using the label names not the images. The filenames ended in XML. So, replaced to PNG filetype. tree_type = tree = ElementTree.parse(file) root = tree.getroot() boxes = list() 'Tree' # get the root of the document # extract eachbounding box The section parses through the xml file. Crates the name of the class. In tree_type. tree = ElementTree.parse(file) parses the file. Then root = tree.getroot() gets the root of the file. And saves bonding box in a list. box root.findall( ):
     xmin = int(box.find( ).text)
     ymin = int(box.find( ).text)
     xmax = int(box.find( ).text)
    ymax = int(box.find( ).text)

   row_list = [filename, tree_type, xmin, xmax, ymin, ymax]
   print(row_list)
   List_dataframe.append(row_list) for in './/bndbox' 'xmin' 'ymin' 'xmax' 'ymax' The for loop goes through each of the boxes and extracts the coordinates. Of the four corners of the box. Then a list is created which saves the filename, class, and coordinates. Then appended to the dataframe from earlier. print(list_dataframe)
	
df = pd.DataFrame(list_dataframe, columns=[ , , , , , ])
	
print(df.head())
	
df.to_csv( , index= ) 'filename' 'tree_type' 'xmin' 'xmax' 'ymin' 'ymax' 'testing.csv' False After that list_dataframe containing all the data of the files. Is then added into another dataframe with the column names. And df.to_csv turns the dataframe into a CSV file. Implementing the model Now the tutorial asks for use to convert the CSV file into txt file. As the model uses a different format. filepath, x1,y1,x2,y2,class_name Where: filepath is the path of the training image x1 is the xmin coordinate for bounding box y1 is the ymin coordinate for bounding box x2 is the xmax coordinate for bounding box y2 is the ymax coordinate for bounding box class_name is the name of the class in that bounding box pandas pd

data = pd.DataFrame()
data[ ] = train[ ] i range(data.shape[ ]):
    data[ ][i] = + data[ ][i] i range(data.shape[ ]):
    data[ ][i] = data[ ][i] + + str(train[ ][i]) + + str(
        train[ ][i]) + + str(train[ ][i]) + + str(train[ ][i]) + + train[ ][i]

data.to_csv( , header= , index= , sep= ) import as 'format' 'image_names' # as the images are in train_images folder, add train_images before the image name for in 0 'format' 'train_images/' 'format' # add xmin, ymin, xmax, ymax and class as per the format required for in 0 'format' 'format' ',' 'xmin' ',' 'ymin' ',' 'xmax' ',' 'ymax' ',' 'cell_type' 'annotate.txt' None None ' ' Now the tutorial asks us clone this When I ran the model I got a few errors. So, I went inside the repository and edited the code. Most errors were AttributeError: module 'keras.backend' has no attribute 'image_dim_ordering'. I found this helpful solving the problem. This is because some of the code used is deprecated. And you need to use the new keras API which has different names for the functions. GitHub repository. GitHub issue answer So, after a few hours of googling other issues the machine learning model started to run. I used a google cloud deep learning VM to train my model. Image by author Due to the amount of time. A decided to reduce the amount epochs to 5. As each epoch took 4 hours. And the default number was 1000. Which will take couple of days to train and quite a bit of money. Now I needed to test model. I did this on my laptop. As Google’s VM did not show images To do this I ran the statement: python test_frcnn.py -p test_images Image by author The output shows it found some trees: Images by author The model was able to find some areas with trees. Mainly thick forestry. But others, not at all: Images by author It looks like the model can’t point individual trees. But it can point to big areas forestry. Improvements Like I mentioned above the model can’t spot individual trees. But areas of forestry. But even then with the areas of forestry. It can only spot some sections. A few small boxes cover a large area of forestry. The dataset was kept brief. So, I can get a workable prototype up and running. Labelling can be improved by boxing every single tree in an image. And having custom segmentation for areas forestry. Instead of boxes to cover an area of forestry. The dataset is very small (20 images) and trained for only 5 epochs. This can be improved by simply adding more images to the dataset. And spending more time training the model. A serious question to ask is, how I would label a large dataset. Labelling 20 images was already long enough. By manually labelling the trees. Took around hour or so. Can’t imagine how to label thousand images by hand. Also, with custom segmentation for large areas forestry I’m not sure how to make custom shapes. To cover the shape of the forestry. As the program I was using only allows for boxes. Not custom shapes like polygons. Published from my website

How To Identify Trees with Deep Learning

This AI is hungry - Part I - Deep Learning Tutorial: Classification of Cooking Dishes and Recipes…

Sentinel Protocol Review — The “RoboCop” for the Wild Wild West of the Crypto Space

Getting to the end of 2022, there’s one soon-to-launch NFT project that has piqued the interest of the industry that everyone is waiting to launch: Valhalla. Whenever situations come up like this, it’s easy to buy into the hype and make poor judgment decisions, I know I sure did when launched last year (thankfully I didn’t end up buying). Mekaverse I decided to do a deep dive into Valhalla and committed to going through their entire Twitter feed, Discord, website, news stories, etc., to best gauge whether this project had the merit to back up the hype that it has garnered up to this point. These are my findings. Project Summary Valhalla is an NFT project focused on being the number one gaming NFT in the world. “Valhalla is a utility-enabled PFP project for gaming and streetwear enthusiasts. Members will be granted access to our exclusive community, events, and digital drops.” The team behind the project is , a studio founded by serial entrepreneur —2x founder and Forbes 30 under 30—that is focused on creating the Web3 version of Twitch. Stacked Studios Alex Lin Y Combinator “We are creating a novel experience for our members, tailormade for those as passionate about gaming and culture as we are.” Stacked raised $13M in its series A funding round led by Pantera Capital (bringing the total to $15M raised). The goal is to turn video streaming creators into shareholders of the platform through the use of governance tokens, but users won’t need to be crypto-savvy to use the service. The model will “align the interests of content operators and creators for the first time.” Stacked is expected to launch in North America in December 2022. https://techcrunch.com/2022/08/31/stacked-web3-game-streaming/?embedable=true Leading up to their mint, Valhalla ran a unique application process. Existing members were giving codes to send friends, so the application pool grew purely by referrals and word of mouth. The model was an incredible success, generating over 11,000 applications! Key Links Twitter Website Discord Mint Details Mint Date TBA Mint Price .25 WETH for presale, subject to change in public sale Mint Type Private Mintlist Sold Out? TBD Notably, Valhalla is running a presale where community members and reputable collectors can reserve an allocation ahead of the public sale. NFTs purchased in the presale will be subject to a 3-mont unlock period. 500 NFTs will be included in the presale and cost .25 WETH. If someone purchases 15 NFTs, 5 will unlock and be able to be sold at the end of each month (equal 1/3 monthly vesting). Collection Details NFTs 10,000 Contract TBA Unique Holders TBD 1 of 1’s? TBA Royalties TBA Team - CEO Alex Lin - Chief of Staff Ryan Wen - SR Software Engineer Victor Cheng - Product Manager Samantha Chow Terry Kang - Software Engineer Roberto Berwa - Software Engineer - Lead Software Engineer Ding Zhang Matt Henderson - Software Engineer Team Experience Y Combinator, LVL6, Hush, Gauge Capital, Microsoft, Open Table, Goldman Sachs, EQUIAM, MIT Media Lab, Greenzie, Vana, Facebook, Uber, Allelo, Atlassian Partners , a pro Valorant player Wardell , the #1 League of Legends team in China JD Gaming , the premier esports team in Asia-Pacific Talon esports , an esports team competing in Valorant, Fortnite, Halo, Wild Rift, and Apex Legends Sentinels , 3x world champions in League of Legends T1 esports Key Lore None available currently Verdict (10/10)⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐ Like and before it, Valhalla is the latest example of a highly qualified team putting in a year+ of work (beginning in 2021 when NFTs exploded) and finally coming to market. Ey3k0n KPR This is a new meta that took time to develop, and it’s one that will ultimately be incredibly beneficial to the NFT space as a whole, as the bar for quality will continue to rise and the influx of top-quality talent will rise alongside it. Despite not knowing the true utility of the NFTs being sold, Valhalla remains a 10/10 for me because it is already funded with years of runway and will predictably integrate well into Stacked, a project I’m incredibly excited to see develop moving forward. When also considering the project has accumulated over 11,000 applications for early access, the gaming partnerships already formed, the team’s background, and the quality of everything that has been publicly unveiled so far, I can’t emphasize enough how positive my outlook is on Valhalla at this point in time. I’m looking forward to their upcoming mint and hope to be a part of it myself!

DEEP SENTINEL

deepsentinel.com

-- employees

Since n.d.

COMPANY RANKING

DEEP SENTINEL

EVERGREEN INDEX #2962

Representing this company? Click here to claim and customize this page!

Deep Sentinel WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!

DEEP SENTINEL

deepsentinel.com

-- employees

Since n.d.

COMPANY RANKING

DEEP SENTINEL

EVERGREEN INDEX #2962

RELATED COMPANIESRANK

STORIES ABOUT DEEP SENTINEL

Representing this company? Click here to claim and customize this page!

Deep Sentinel WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!