Debaditya Roy
"Always do your best, whatever you plant now will harvest later" - Og Mandino
Email - roy_debaditya at ihpc dot a-star dot edu dot sg, debadityaroy5555 at gmail dot com
I am currently a Scientist at the Institute of High-Performance Computing, A*STAR, Singapore working on Action Anticipation, Situated Reasoning and Neurosymbolic AI with Dr. Basura Fernando.
Previously, I was a post-doctoral researcher working with Prof. Atsushi Fukuda in the Department of Transportation Engineering at Nihon University, Japan in the M2Smart project funded by JICA and JST.
I received my doctoral degree from the Department of Computer Science and Engineering at the Indian Institute of Technology Hyderabad as a member of the Visual Learning and Intelligence (VIGIL) lab. I was advised by Prof. C. Krishna Mohan and my thesis was titled Representation Learning for Action Recognition. My work on snatch theft detection was featured in The Hindu, a leading daily newspaper in India.
News
Our paper titled "Predicting the Next Action by Modeling the Abstract Goal" is accepted to ICPR 2024.
Our paper titled "TSANet:Forecasting Traffic Congestion Patterns from Aerial Videos using Graphs and Transformers" is accepted to Pattern Recogntion journal. Congrats Naveen!
Serving as reviewer for ECCV 2024, Nature Scientific Reports
Patent granted on "Method and System for Detection of Crime Events in Surveillance Videos"[pdf]
Two papers accepted at WACV 2024
Debaditya Roy, Ramanathan Rajendiran, and Basura Fernando, "Interaction Visual Transformer for Egocentric Action Anticipation", ranked #1 on EPIC-KITCHENS 100 dataset [pdf] [code]
Debaditya Roy, Dhruv Verma, and Basura Fernando, "ClipSitu: Effectively Leveraging CLIP for Conditional Predictions in Situation Recognition" State of the art on imSitu State of the art on SWiG [pdf][code]
Serving as reviewer for CVPR 2024
Serving as reviewer for WACV 2024
Our team is ranked #1 on EPIC-KITCHENS 55 dataset
Debaditya Roy and Basura Fernando"Predicting the Next Action by Modeling the Abstract Goal", [pdf]
Our paper titled "Improving Multi-agent Trajectory Prediction using Traffic States on Interactive Driving Scenarios"[pdf] is accepted in IEEE Robotics and Automation Letters. Congratulations Vishnu and Abhinav
Serving as reviewer for ICCV 2023.
Granted patent titled "A method and system for real-time detection of traffic violation by two-wheeled riders" from Indian Patent Office. Congratulations to all my co-authors.
Granted patent titled "Method and system for detection of accident in traffic surveillance video" from Indian Patent Office. Congratulations to all my co-authors.
Gave an invited Talk - "Predicting Human Behavior - From the Kitchen to the Roads" in INCAPS - 2nd Indo-Norway Workshop on Smart Sensing, Communication and Machine Learning for Autonomous and Cyber Physical Systems conducted at IIT Hyderabad
New preprint - "Predicting the Next Action by Modeling the Abstract Goal"