Research

Computer Sciences and Information Technology

Title :

Manageable AI for scene understanding tasks from low light noisy images

Area of research :

Computer Sciences and Information Technology

Focus area :

Artificial Intelligence

Principal Investigator :

Dr. Jhilik, Thapar Institute of Engineering and technology, Patiala, Punjab (147004)

Contact info :

Details

Executive Summary :

Image and video acquisition and processing are crucial components of ICT, influencing consumer behavior, products, and services. This project aims to create manageable AI equipment that can be used with limited computational and energy resources, especially in mobile contexts. The main challenge is image understanding in challenging environments, such as low light or backlit environments with camera noise or poor resolution. The proposed project focuses on three issues: image understanding in challenging environmental conditions, developing lightweight versions of scene understanding techniques for real-time tasks on mobile devices, and generating interpretable methods and models for Explainable AI XAI tools. The project will be executed in three steps: Studying low light image enhancement algorithms, applying suitable enhancement operators, and developing lightweight versions of scene understanding model enhancement detection. The Indian and Italian partners will work together to devise lightweight network architectures that allow at least partial interpretability. The project aims to improve the accuracy of scene understanding models in real-time softwares on mobile devices.

Total Budget (INR):

4,50,000

Organizations involved