|
Advanced Applications Development Facility |
ARTIFICIAL INTELLIGENCE based application development
The activities of Computer Division, RRCAT in the domain of Artificial Intelligence and Machine Learning comprise of building and exploration of advanced applications for spatial, temporal and spatio-temporal data. This involves exploring the issues in training a Deep Convolutional Neural Network on images (spatial data) and suggesting insights and remedies in order to overcome them.
An advanced vehicle counting application was developed, which counts moving vehicles in surveillance videos (temporal-spatial data). For object detection start-of-the-art object detection algorithms have been applied. For counting number of vehicles, two robust counting algorithms have been developed.
For development of applications for automated diagnostics of ailments using medical imaging, intelligent agent based on Deep Convolutional Neural Networks has been developed or automated detection of Diabetic Retinopathy (DR) using colour fundus images of retina.
Intelligent Agent for Counting and Detection of Moving Vehicles in Videos
Intelligent agent has been developed for detection and counting of moving vehicles in videos using Deep Learning based algorithms. Convolutional Neural Network (CNN) and publically available COCO dataset is used to develop and train the agent. The agent is tested over variety of videos and produced near 100% counting accuracy.
Detection and counting of moving vehicles in video
Intelligent web based software for automated detection of Diabetic Retinopathy using Deep Learning
One web based software is developed for screening of patients suffering from Diabetic Retinopathy (DR). Artificial intelligence based agent has been designed and developed for automated detection of Diabetic Retinopathy (DR) using deep learning methods. Algorithms are developed for pre-processing and classification of colour fundus images of retina. Deep learning models have been developed and trained using publicly available dataset, EyePACS, which contains 75,000 images. Results of developed models are very encouraging - Area Under Receiver Operating Characteristic (AUROC/AUC) - obtained as 0.92. This software will act as an aid to the manual diagnostic process by referring DR patients to an ophthalmologist for further examination (if detected positive) well in time, thus reducing the risks of vision loss.
Intelligent software for detection of Diabetic Retinopathy in colour fundus images of retina
|
|
