About Me

I am a research scientist at Facebook AI Research (FAIR) working on computer vision and machine learning. My current research focuses on modeling temporal dynamics, with applications including video understanding and physical reasoning. I obtained a PhD from Carnegie Mellon University (CMU) where I worked with Deva Ramanan (here's a link to my dissertation). Earlier I graduated with a masters from CMU as well, working with Martial Hebert, Abhinav Gupta, Kris Kitani and David Fouhey as a Siebel Scholar. Even before I was a CS undergrad at IIIT, Hyderabad, working with C. V. Jawahar. I have also been fortunate to work with some amazing people through internships, at DeepMind (with Andrew Zisserman, João Carreira and Carl Doersch), Adobe Research (with Josef Sivic and Bryan Russell) and Facebook AI (with Lorenzo Torresani, Georgia Gkioxari and Du Tran).

Preprints

Forward Prediction for Physical Reasoning

Rohit Girdhar, Laura Gustafson, Aaron Adcock and Laurens van der Maaten
arXiv 2020 · pdf

Publications

CATER: A diagnostic dataset for Compositional Actions and TEmporal Reasoning

Rohit Girdhar and Deva Ramanan
ICLR 2020 (oral) · Addis Ababa, Ethiopia · webpage
Holistic Video Understanding (HVU) workshop, ICCV 2019 (oral) · Seoul, South Korea
Best paper award at HVU Workshop, ICCV 2019

MetaPix: Few-Shot Video Retargeting

Jessica Lee, Deva Ramanan and Rohit Girdhar
ICLR 2020 · Addis Ababa, Ethiopia · webpage
MetaLearn workshop, NeurIPS 2019 (oral) · Vancouver, Canada
One of top-2 papers (out of 84 submissions) selected for full oral at MetaLearn, NeurIPS'19

Are we asking the right questions in MovieQA?

Bhavan Jasani, Rohit Girdhar and Deva Ramanan
Closing the Loop Between Vision and Language (CLVL) workshop, ICCV 2019 (spotlight) · Seoul, South Korea · webpage

Video Action Transformer Network

Rohit Girdhar, João Carreira, Carl Doersch and Andrew Zisserman
CVPR 2019 (oral) · Long Beach, CA · webpage

DistInit: Learning Video Representations without a Single Labeled Video

Rohit Girdhar, Du Tran, Lorenzo Torresani and Deva Ramanan
ICCV 2019 · Seoul, South Korea · pdf
Learning from Unlabled Videos (LUV) Workshop, CVPR 2019 · Long Beach, CA · pdf

A Better Baseline for AVA

Rohit Girdhar, João Carreira, Carl Doersch and Andrew Zisserman
ActivityNet Workshop, CVPR 2018 (oral) · Salt Lake City, UT · pdf
Close second in AVA action recognition challenge

Detect-and-Track: Efficient Pose Estimation in Videos

Rohit Girdhar, Georgia Gkioxari, Lorenzo Torresani, Manohar Paluri and Du Tran
CVPR 2018 · Salt Lake City, UT · webpage

Simple, efficient and effective keypoint tracking

Rohit Girdhar, Georgia Gkioxari, Lorenzo Torresani, Deva Ramanan, Manohar Paluri and Du Tran
PoseTrack Workshop, ICCV 2017 (oral) · Venice, Italy · pdf
First in keypoint tracking challenge

Attentional Pooling for Action Recognition

Rohit Girdhar and Deva Ramanan
NeurIPS 2017 · Long Beach, CA · webpage

ActionVLAD: Learning spatio-temporal aggregation for action classification

Rohit Girdhar, Deva Ramanan, Abhinav Gupta, Josef Sivic and Bryan Russell
CVPR 2017 · Honolulu, HI · webpage

Binge Watching: Scaling Affordance Learning from Sitcoms

Xiaolong Wang*, Rohit Girdhar* and Abhinav Gupta (* equal contribution)
CVPR 2017 (spotlight) · Honolulu, HI · webpage · pdf · data

Learning a Predictable and Generative Representation for Objects

Rohit Girdhar, David Fouhey, Mikel Rodriguez and Abhinav Gupta
ECCV 2016 (spotlight) · Amsterdam, Netherlands · webpage

Cutting through the clutter: Task-relevant features for image matching

Rohit Girdhar, David Fouhey, Kris Kitani, Abhinav Gupta and Martial Hebert
WACV 2016 · Lake Placid, NY · pdf

Optimizing Storage Intensive Vision Applications to Device Capacity

Rohit Girdhar, Jayaguru Panda and C. V. Jawahar
ACCV 2014 · Singapore · pdf

Posts

Dec 23, 2016 · Compile TensorFlow on CentOS 6

Fun Stuff

Inspired by the amazing work of David Fouhey, I have dabbled in the fine art of joke publications. Here's a taste.

PSYCHO: PerSonalitY CHaracterizatiOn of artificial intelligence

Achal Dave and Rohit Girdhar
SIGBOVIK 2018 · Pittsburgh, PA · pdf