Research
I have always been excited about building intelligent systems that work. I draw inspiration from natural sciences, and my ideas revolve around computer vision, machine learning, robotics and, systems. Currently, I'm learning about the geometric and physical structure of our world to augment next generations of computational intelligence.
|
|
ROADWork Dataset: Learning to Recognize, Observe, Analyze and Drive Through Work Zones
Anurag Ghosh, Robert Tamburo, Shen Zheng, Juan R. Alvarez Padilla, Hailiang Zhu, Michael Cardei, Nicholas Dunn, Christoph Mertz, Srinivasa Narasimhan
[Website][Paper][Code]
Largest open-source dataset for studying autonomous driving in work zones.
|
|
Saliency Guided Image Warping for Unsupervised Domain Adaptation
Shen Zheng★, Anurag Ghosh★, Srinivasa Narasimhan
[Website][Paper][Code]
Oversample salient object regions by warping source-domain images in-place during training while performing domain adaptation. Improve adaptation across geographies, lighting and weather conditions, is agnostic to the task, domain adaptation algorithm, saliency guidance, and underlying model architecture.
|
|
Learned Two-Plane Perspective Prior based Image Resampling for Efficient Object Detection
Anurag Ghosh, N Dinesh Reddy, Christoph Mertz, Srinivasa Narasimhan
Computer Vision and Pattern Recognition Conference(CVPR), 2023
[Website][Paper][Code]
A learnable geometry-guided prior that incorporates rough geometry of the 3D scene (a ground plane and a plane above) to resample the images for efficient object detection. This significantly improves small and far-away object detection performance while also being more efficient.
|
|
Exploiting Tradeoffs in Resource Constrained Vision
Novel methods for improving trade-offs at different levels of vision system abstractions. We considered constraints like real-timeness, network latency/bandwidth, inter-process contention, onboard compute, power, heat and battery considerations.
Relevant Publications
[1] Chanakya: Learning Runtime Decisions for Adaptive Real-Time Perception
Anurag Ghosh, Vaibhav Balloli, Akshay Nambi, Aditya Singh, Tanuja Ganu
Conference on Neural Information Processing Systems(NeurIPS), 2023
[Website][Paper][Code]
Honorable Mention, Streaming Perception Challenge, CVPR 2021. [Presentation at WAD]
[2] REACT: Streaming Video Analytics On The Edge With Asynchronous Cloud Support
Anurag Ghosh, Srinivasan Iyengar, Stephen Lee, Anuj Rathore, Venkat Padmanabhan
International Conference on Internet of Things Design and Implementation (IoTDI), 2023
[Paper]
[3] Holistic Energy Awareness for Intelligent Drones
Srinivasan Iyengar, Ravi Raj Saxena, Joydeep Pal, Bhawana Chhaglani, Anurag Ghosh, Venkat Padmanabhan, Prabhakar T. Venkata
International Conference on Systems for Energy-Efficient Built Environments (BuildSys), 2021
[Paper]
(Best Paper Runner-Up)
(Also appeared in Transactions on Sensor Networks)
|
|
Smartphone-based Driver License Testing
Watch Microsoft CEO Satya Nadella explain the project!
Read about our work on PM Awards Innovations Coffee Table Book! (Extracted here)
Deployed in multiple states/10+ cities in India, automatically testing hundreds of thousands of drivers at a low-cost with >99% accuracy (test verified by human operator). See Overview and Dashboard.
Instead of prohibitively expensive (>150,000$) overhead pole-mounted camera infrastructure to estimate car trajectories and judge driver manuevers, we use sub-500$ smartphones and additionally get driver state monitoring for free (face verification, mirror scanning, distraction and seatbelt checks).
Relevant Publications
[1] Smartphone-based Driver License Testing
Anurag Ghosh, Vijay Lingam, Ishit Mehta, Akshay Nambi, Venkat Padmanabhan, Satish Sangameswaran, Conference on Embedded Networked Sensor Systems (SenSys Demo), 2019
[Paper]
[2] ALT: Towards Automating Driver License Testing using Smartphones
Akshay Nambi, Ishit Mehta, Anurag Ghosh, Vijay Lingam, Venkat Padmanabhan, Conference on Embedded Networked Sensor Systems (SenSys), 2019
[Paper]
|
|
Analyzing Racket Sports From Broadcast Videos
Piloted with ESPN/Star Sports at Premier Badminton League, watched by tens of millions in South East Asia.
Do we really need half a million dollars for HawkEye to understand players? Our End-to-end framework automatically tags broadcast sport videos in near-real time. Our analysis shows a single camera suffices for mining rich and actionable player data, instead of relying on existing cumbursome multi-camera setups or sensors. It is used for live broadcast visualizations.
Relevant Publications
[1] Analyzing Racket Sports From Broadcast Videos
Anurag Ghosh, IIIT Hyderabad (Master's Thesis), 2019
[Paper]
[2] Towards Structured Analysis of Broadcast Badminton Videos
Anurag Ghosh, Suriya Singh, C.V. Jawahar, Winter Conference On Applications of Computer Vision (WACV), 2018
[Paper]
[3] SmartTennisTV: An automatic indexing system for tennis
Anurag Ghosh, C.V. Jawahar, National Conference on Computer Vision, Pattern Recognition, Image Processing and Graphics (NCVPRIPG), 2017
[Paper]
(Best Paper Award)
|
|
Signals Matter: Understanding Popularity and Impact on Stack Overflow
Arpit Merchant, Daksh Shah, Gurpreet Singh Bhatia, Anurag Ghosh, Ponnurangam Kumaraguru
The Web Conference (WWW), 2019
[Paper]
|
|
Dynamic narratives for heritage tour
Anurag Ghosh ★, Yash Patel ★, Mohak Sukhwani, C.V. Jawahar
VisArt Workshop, Europen Conference on Computer Vision (ECCV), 2016
[Paper]
|
Interesting/Inspiring Links
|
|
|
|
|
|
|
|
|