Computer Sciences and Information Technology
Title : | Development of an Intelligent Visual Query Response Solution for Remote Sensing |
Area of research : | Computer Sciences and Information Technology |
Principal Investigator : | Dr. Suman Kumar Maji, Indian Institute Of Technology (IIT) Patna, Bihar |
Timeline Start Year : | 2023 |
Timeline End Year : | 2026 |
Contact info : | suman.maji@gmail.com |
Equipments : | GPU
Laptop cooling pad
Laptop |
Details
Executive Summary : | Remote sensing is the process of observing and collecting specific information, typically in the form of images, over an area from a distance. Interpreting these images is challenging due to the abundance of useful and critical information and requires expert technical knowledge. This project aims to develop an artificial intelligent (AI) model that can correctly answer questions in natural language regarding an input RSI. The proposed system consists of five basic sequential modules: Denoising module, Input module, Representation module, Fusion module, and Answering module. The denoising module improves the spatial quality of the images, while the input module contains denoised visuals and respective questions. The representation module extracts various independent features from the given inputs, while the fusion module integrates these features into a vectorized encoding of useful features. The answering module uses a multilayer perceptron classifier to decode the desired output from the encoded feature set. Understanding an image the human way is a challenging task, similar to designing a system as intelligent as a human. This problem has been widely accepted as AI-complete and suggested as a Visual Turing Test. Designing such a system will pave the pathway for designing a remote sensing visual search engine that understands RSI and adapts to different types of requests, making it a significant breakthrough. |
Co-PI: | Dr. Somanath Tripathy, Indian Institute Of Technology (IIT) Patna, Bihar-801106 |
Total Budget (INR): | 26,67,981 |
Patents : | 1 |
Organizations involved