Hello and Welcome!
I work on problems relating to Multimodal Understanding: Object Segmentation, Multiple Object Tracking, Action Localization/Detection. Broadly, I am interested in solving real-life problems (agriculture, autonomous driving, medicine, drug discovery) using discrete / continuous optimization, deep learning, and machine learning. I also like to work on computer graphics and image processing problems.
Please refer to my Projects page for details on current and previous works of mine.
Academic Profile:
I defended my PhD (Dec - 2014) on Video Understanding (Unsupervised Segmentation, Metric Learning) at STARS TEAM, INRIA, France (Thesis). [Supervised by Guillaume Charpiat and Monique Thonnat]
I have also had the opportunity to collaborate with Mike Jones, Dhruv Batra, and Tim Marks.
I hold MS in ECE from University of Florida and BS in ECE from Manipal University.
Community Services and Acknowledgments:
2023: 1000+ citations! in Google Scholar.
Awarded outstanding reviewer! for CVPR.
Program Committee/Reviewer: REID-MTMC'17, Prediction in Wild 18, AIC, WACV'19 '20,
IPAS'18, CCIP'15, IJCNN'19-24. TIP, MVA, CVPR'20 '21 '22 '23' 24, ECCV' 22, '24 ICCV '19 '21 '23
Talks/Tutorials:
Invited lecture at Georgia Tech on Visual embedding and advancements.
Invited keynote talk on Multi-Camera Object Tracking & Deep Metric Learning at IEEE IPAS.
I authored a chapter on Visual Embeddings in Deep Learning for Vision Systems.
Invited tutorial at IEEE Silicon Valley Chapter, on Re-identification in Camera Networks and Deep Learning.
News:
Patent Granted! (Pending patents are in my Scholar Page)
@Argo/Latitude AI - Auto-magically produce Amodal (true extent) cuboids from Modal Labels.
@NVIDIA - Vehicle re-identification using deep learning with low latency.
@NVIDIA - Smart Garage using AI (I led the perception architecture).
@NVIDIA - Associating stuffs to owners in multi-person scene using deep learning (real time!).
@NVIDIA - Simulation Engine for Smart Garage Environments.
Argoverse-2 accepted! in Neurips 2021.
CVPR 2019 Paper accepted for Oral! session (top 5%).
IJCNN 2019 Oral! Paper acceptance.
PAMTRI (How to effectively mix synthetic and real data for training deep nets) accepted in ICCV 2019.