Object Detection with Tracking

Object detection is used to identify multiple relevant objects in a single image. In broader sense, it provides localization of the objects in an image. Different algorithms such as YOLO, Faster R-CNN, Mask R-CNN, NASNet, SSD are used to implement such. Have a look on video entitled as Object Detection Custom Training of Image Mask RCNN Deep Learning for Mask R-CNN and Testing custom Object detector with tensorflow object detection api for Faster R-CNN.


Human Pose Estimation

A Human Pose Skeleton represents the orientation of a person in a graphical format. Essentially, it is a set of coordinates that can be connected to describe the pose of the person. For multiple persons in an image either top-down or down-top approach is applied. Let us draw some light on its use case. One could be monitoring the activities of person (it will cover health status). Second would be in training of robots. We as a team has prepared a demo for this which one can see by clicking Real time Pose Estimation in Video using Posenet using Deep Learning.


Optical Character Recognition

It is the process of mechanically or electronically converting scanned images of handwritten, typed or printed text into machine-encoded text. Both handwritten and printed characters can be recognized and converted into machine readable text. Each character is segmented first using the appropriate method such as contours. Each character is labelled and character images along with the labels are trained using convolutional neural network.Trained Model is deployed for the unknown text in an image. For demo, oe can refer to Optical Character Recognition using Python


Image Semantic Segmentation

The image semantic segmentation challenge consists in classifying each pixel of an image (or just several ones) into an instance, each instance (or category) corresponding to an object or a part of the image. Object detection do not provide the full comprehension of a scene. The state-of-the-art models (FCN, ParseNet, FPN, Mask R-CNN, ) use architectures trying to link different parts of the image in order to understand the relations between the objects. One of the demo for this task can be found at Object Detection Custom Training of Image Mask RCNN Deep Learning


Real Time Face Recognition

Let us straightforward plunge into reason why real time face recognition is important automatically. Consider the case one is entering the crucial areas such as banking or any other and if each person is automatically detected by some hardware and checked for person details in the database it will help to know the records of person. COnsider second case where child trafficking is at peak. Such children can be located using this app. In this way deep learning is helping society in some way. AI Sangam has made a demo regarding this hot burning topic which one can view by clicking here.


Image Classification

This is one of the domains where at some point deep learning has surpassed human intelligence. There is difference between image classification, object detection and segmentation. Biggest success in this domain is due to convolutional neural networks and many standard models such as LeNet, AlexNet, VGG, GoogLeNet, ResNet are based on CNN anh has achieved great success in classification. We as a team has been actively working in this domain. There are some demos which we have made and one of them can be explored and watched by clicking here.