Before you go, check out these stories!

Hackernoon logoMajor Image Recognition And Annotation Trends by@koranand

Major Image Recognition And Annotation Trends

Author profile picture

@koranandAnna Korotkova

Image recognition and annotation technologies are evolving. New techniques that allow you to solve a wide variety of tasks quickly appear. We are happy to present five major trends in image recognition and annotation.

Active learning for machine learning

Most supervised machine learning models require huge amounts of data to be trained with good results. And most companies have difficulties with providing data scientists with this data, in particular labeled one.In most cases, data scientists are provided with a big, unlabelled data sets and are asked to train well-performing models with them. In most cases, the amount of data is too big to be manually labeled, and it is quite difficult to train good supervised models with this data.Active learning is the process of prioritising the data which needs to be labeled to have the highest impact on training a supervised model. Active learning can be used in situations where the amount of data is too large to be labeled and some priority needs to be made to label the data in a smart way.

Edge computing

Edge computing is a computing model in which computing takes place near the physical location where data is being collected and analyzed, rather than on a centralized server or in the cloud. This new infrastructure involves sensors to collect data and edge servers to securely process data in real-time on-site, while also connecting other devices, such as laptops and smartphones, to the network.Cloud computing and the internet of things (IoT) have elevated the role of edge devices, ushering in the need for more intelligence, computing power, and advanced services at the network edge. This concept, where processes are decentralized and occur in a more logical physical location, is referred to as edge computing.

Classification+Location & Object detection VS Semantic segmentation & Instance segmentation

Classification+Location: It is possible to classify an image as a cat. The question is if we can get the location of the said cat in that image by drawing a bounding box around the cat? Here we assume that there is a fixed number of objects(commonly 1) in the image. 

Object detection: In a real-world setting, we don’t know how many objects are in the image beforehand. So object detection technology tries to detect all of them.

Semantic segmentation: Given an image, can we classify each pixel as belonging to a particular class? YES. If put together, semantic and instance segmentation methods become a powerful tool. It is possible to both detect object pixels and determine the object location in the image. Semantic instance segmentation is a useful tool for land cover classification, which has various applications. Land mapping via satellite imagery can be useful for governmental institutions to monitor deforestation (especially illegal), urbanization, traffic, and more.

Using non-standard imagery

Computer vision is used to solve a wide variety of tasks. Therefore, images may not always belong to a particular class. Thus, the coronavirus pandemic has had a strong impact on many industries. In computer vision, this trend is reflected in its use for medical tasks. For example, special datasets are now actively created  for machine learning.

Object recognition with point cloud

A point cloud is a collection of data points defined within a three-dimensional coordinate system.This technology is typically used within a space (for example a room or a container) where the location and shape of each object are represented by a list of coordinates (X, Y, and Z). The list of coordinates is referred to as a “point cloud”.This technology provides an accurate representation of where an object is in the space, and any movement can be accurately tracked.

Previously published at


Join Hacker Noon

Create your free account to unlock your custom reading experience.