Rohit Girdhar

Rohit Girdhar

Research Scientist

Meta Superintelligence Labs

I am a Research Scientist at the Meta Superintelligence Labs. My current research focuses on understanding and generating multimodal data. I obtained a PhD from Carnegie Mellon University (here’s a link to my dissertation), where I worked on learning from and understanding videos. I was previously part of the Facebook AI Research (FAIR) group at Meta, and have spent time at DeepMind, Adobe and Facebook as an intern. See here for a formal bio.

News

Education

  • PhD in Robotics, 2019

    Carnegie Mellon University, Pittsburgh PA

  • MS in Robotics, 2016

    Carnegie Mellon University, Pittsburgh PA

  • B. Tech. in Computer Science, 2014

    IIIT Hyderabad, India

Experience

  • Meta · Research Scientist

    New York · 2019 -- Present

  • DeepMind · Research Scientist Intern

    London · Summer 2018

  • Facebook · Research Scientist Intern

    Menlo Park · Summer 2017

  • Adobe · Research Scientist Intern

    San Francisco · Summer 2016

  • Facebook · Software Engineering Intern

    Menlo Park · Summer 2013

Highlights

Videos powered by MovieGen and Emu Video!


Projects and Publications

.js-id-selected
Selected Generative Multimodal Video 3D Representation Detection All
Kumar Ashutosh, XuDong Wang, Xi Yin, Kristen Grauman, Adam Polyak, Ishan Misra, Rohit Girdhar
January, 2026 In arXiv, 2026
Human detectors are surprisingly powerful reward models
Human detectors are surprisingly powerful reward models

Using human detection confidence as a simple yet effective reward model to improve human motion in video generation.
PDF Cite Video
Bolin Lai, XuDong Wang, Saketh Rambhatla, James M. Rehg, Zsolt Kira, Rohit Girdhar, Ishan Misra
November, 2025 In CVPR, 2026
Toward Diffusible High-Dimensional Latent Spaces: A Frequency Perspective
Toward Diffusible High-Dimensional Latent Spaces: A Frequency Perspective

Introducing FreqWarm, a plug-and-play frequency warm-up curriculum that improves high-dimensional latent diffusion by increasing early-stage exposure to high-frequency signals.
PDF Cite
Yinbo Chen, Rohit Girdhar, Xiaolong Wang, Sai Saketh Rambhatla, Ishan Misra
January, 2025 In arXiv, 2025
Diffusion Autoencoders are Scalable Image Tokenizers
Diffusion Autoencoders are Scalable Image Tokenizers

Simplified image tokenization using diffusion
PDF Cite Code
Kumar Ashutosh, Yossi Gandelsman, Xinlei Chen, Ishan Misra, Rohit Girdhar
January, 2025 In ICML, 2025
LLMs can see and hear without any training
LLMs can see and hear without any training

Pure text-only LLMs can use off-the-shelf multimodal embedding models to do various multimodal tasks!
PDF Cite Code
Shijie Wang, Samaneh Azadi, Rohit Girdhar, Saketh Rambhatla, Chen Sun, Xi Yin
December, 2024 In CVPR, 2025
MotiF: Making Text Count in Image Animation with Motion Focal Loss
MotiF: Making Text Count in Image Animation with Motion Focal Loss

Using flow to improve motion in video generation
PDF Cite
MovieGen Team (Core-Contributor)
October, 2024 In arXiv, 2024
Movie Gen: A Cast of Media Foundation Models
Movie Gen: A Cast of Media Foundation Models

State-of-the-Art Video (+Audio) Generation Model
PDF Cite Video
Llama3 Team (Co-Lead the Video Recognition Efforts)
July, 2024 In arXiv, 2024
The Llama 3 Herd of Models
The Llama 3 Herd of Models

State-of-the-Art open-source LLM with multimodal capabilities
PDF Cite Code
Changan Chen, Ashutosh Kumar, Rohit Girdhar, David Harwath, Kristen Grauman
April, 2024 In CVPR, 2024
SoundingActions: Learning How Actions Sound from Narrated Egocentric Videos
SoundingActions: Learning How Actions Sound from Narrated Egocentric Videos

Self-supervised embedding to learn how actions sound from narrated in-the-wild egocentric videos.
PDF Cite
XuDong Wang, Trevor Darrell, Sai Saketh Rambhatla, Rohit Girdhar, Ishan Misra
February, 2024 In CVPR, 2024
InstanceDiffusion: Instance-level Control for Image Generation
InstanceDiffusion: Instance-level Control for Image Generation

SOTA instance-conditioned diffusion model for image generation.
PDF Cite Code
Sachit Menon, Ishan Misra, Rohit Girdhar
December, 2023 In CVPR, 2024
Generating Illustrated Instructions
Generating Illustrated Instructions

Introducing a new task of generating instructions for the task you want to solve with illustrations, and a LLM + Diffusion model based solution.
PDF Cite Code
Wilson Yan, Andrew Brown, Pieter Abbeel, Rohit Girdhar, Samaneh Azadi
November, 2023 In arXiv, 2023
Motion-Conditioned Image Animation for Video Editing
Motion-Conditioned Image Animation for Video Editing

Image animation FTW again! SOTA video editing results by animating the first frame with motion conditioning.
PDF Cite
XuDong Wang, Ishan Misra, Ziyun Zeng, Rohit Girdhar, Trevor Darrell
November, 2023 In CVPR, 2024
VideoCutLER: Surprisingly Simple Unsupervised Video Instance Segmentation
VideoCutLER: Surprisingly Simple Unsupervised Video Instance Segmentation

SOTA unsupervised video segmentation using CutLER.
PDF Cite Code
Rohit Girdhar, Mannat Singh, Andrew Brown, Quentin Duval, Samaneh Azadi, Sai Saketh Rambhatla, Akbar Shah, Xi Yin, Devi Parikh, Ishan Misra
November, 2023 In ECCV, 2024
Emu Video: Factorizing Text-to-Video Generation by Explicit Image Conditioning
Emu Video: Factorizing Text-to-Video Generation by Explicit Image Conditioning

A simple and effective approach to high-quality video generation by learning to animate high quality images.
PDF Cite Demo
Rohit Girdhar, Alaaeldin El-Nouby, Zhuang Liu, Mannat Singh, Kalyan Vasudev Alwala, Armand Joulin, Ishan Misra
May, 2023 In CVPR, 2023 (Highlighted Presentation)
ImageBind: One Embedding Space To Bind Them All
ImageBind: One Embedding Space To Bind Them All

One embedding space for 6 different modalities, enables zero-shot recognition on all modalities!
PDF Cite Video Code
Mannat Singh, Quentin Duval, Kalyan Vasudev Alwala, Haoqi Fan, Vaibhav Aggarwal, Aaron Adcock, Armand Joulin, Piotr Dollár, Christoph Feichtenhofer, Ross Girshick, Rohit Girdhar, Ishan Misra
March, 2023 In ICCV, 2023
The effectiveness of MAE pre-pretraining for billion-scale pretraining
The effectiveness of MAE pre-pretraining for billion-scale pretraining

Scaling up MAE pre-pretraining, followed by weakly supervised pretraining, leads to strong representations.
PDF Cite Code
XuDong Wang, Rohit Girdhar, Stella X. Yu, Ishan Misra
January, 2023 In CVPR, 2023
CutLER: Cut and Learn for Unsupervised Object Detection and Instance Segmentation
CutLER: Cut and Learn for Unsupervised Object Detection and Instance Segmentation

Discovering objects using DINO features, and learning an unsupervised detection + segmentation model
PDF Cite Code
Kumar Ashutosh, Rohit Girdhar, Lorenzo Torresani, Kristen Grauman
January, 2023 In CVPR, 2023 (Highlighted Presentation)
HierVL: Learning Hierarchical Video-Language Embeddings
HierVL: Learning Hierarchical Video-Language Embeddings

Video-language embeddings are a promising avenue for injecting semantics into visual representations, but existing methods capture only short-term associations between seconds-long video clips and their accompanying text. We propose HierVL, a novel hierarchical video-language embedding that simultaneously accounts for both long-term and short-term associations. As training data, we take videos accompanied by timestamped text descriptions of human actions, together with a high-level text summary of the activity throughout the long video (as are available in Ego4D). We introduce a hierarchical contrastive training objective that encourages text-visual alignment at both the clip level and video level. While the clip-level constraints use the step-by-step descriptions to capture what is happening in that instant, the video-level constraints use the summary text to capture why it is happening, i.e., the broader context for the activity and the intent of the actor. Our hierarchical scheme yields a clip representation that outperforms its single-level counterpart as well as a long-term video representation that achieves SotA results on tasks requiring long-term video modeling. HierVL successfully transfers to multiple challenging downstream tasks (in EPIC-KITCHENS-100, Charades-Ego, HowTo100M) in both zero-shot and fine-tuned settings.
PDF Cite Video Code
Yue Zhao, Ishan Misra, Philipp Krähenbühl, Rohit Girdhar
December, 2022 In CVPR, 2023 (Highlighted Presentation)
Learning Video Representations from Large Language Models
Learning Video Representations from Large Language Models

Leveraging LLMs to auto-annotate videos for representation learning.
PDF Cite Colab Code
Rohit Girdhar, Alaaeldin El-Nouby, Mannat Singh, Kalyan Vasudev Alwala, Armand Joulin, Ishan Misra
June, 2022 In CVPR, 2023
OmniMAE: Single Model Masked Pretraining on Images and Videos
OmniMAE: Single Model Masked Pretraining on Images and Videos

Single self-supervised representation for images and videos.
PDF Cite Video Code
Rohit Girdhar, Mannat Singh, Nikhila Ravi, Laurens Van Der Maaten, Armand Joulin, Ishan Misra
June, 2022 In CVPR, 2022 (Oral Presentation)
Omnivore: A Single Model for Many Visual Modalities
Omnivore: A Single Model for Many Visual Modalities

A single model for images, video and single-view 3D.
PDF Cite Code
Kristen Grauman, Andrew Westbury, Rohit Girdhar, Et Al
March, 2022 In CVPR, 2022 (Best paper finalist)
Ego4D: Around the World in 3,000 Hours of Egocentric Video
Ego4D: Around the World in 3,000 Hours of Egocentric Video

The largest egocentric video dataset.
PDF Cite Video Code
Xingyi Zhou, Rohit Girdhar, Armand Joulin, Philipp Krähenbühl, Ishan Misra
January, 2022 In ECCV, 2022
Detecting Twenty-thousand Classes using Image-level Supervision
Detecting Twenty-thousand Classes using Image-level Supervision

Leverages image classification data to build an object detector
PDF Cite Colab Code
Bowen Cheng, Anwesa Choudhuri, Ishan Misra, Alexander Kirillov, Rohit Girdhar, Alexander G. Schwing
December, 2021 In arXiv, 2021
Mask2Former for Video Instance Segmentation
Mask2Former for Video Instance Segmentation

SOTA video segmentation using Mask2Former.
PDF Cite Code
Bowen Cheng, Ishan Misra, Alexander G. Schwing, Alexander Kirillov, Rohit Girdhar
December, 2021 In CVPR, 2022
Masked-attention Mask Transformer for Universal Image Segmentation
Masked-attention Mask Transformer for Universal Image Segmentation

Single architecture state-of-the-art in instance, semantic and panoptic segmentation.
PDF Cite Code
Ishan Misra, Rohit Girdhar, Armand Joulin
September, 2021 In ICCV, 2021 (Oral Presentation)
3DETR: An End-to-End Transformer Model for 3D Object Detection
3DETR: An End-to-End Transformer Model for 3D Object Detection

First Transformer based detection architecture for 3D data.
PDF Cite Code
Rohit Girdhar, Kristen Grauman
June, 2021 In ICCV, 2021
Anticipative Video Transformer
Anticipative Video Transformer

An autoregressive video transformer architecture for action anticipation in videos.
PDF Cite Code
Zhongzheng Ren, Ishan Misra, Alexander G. Schwing, Rohit Girdhar
May, 2021 In CVPR, 2021
3D Spatial Recognition without Spatially Labeled 3D
3D Spatial Recognition without Spatially Labeled 3D

WyPR can detect and segment objects in a 3D scene without needing any spatial labels at all!
PDF Cite Slides Code
Zaiwei Zhang, Rohit Girdhar, Armand Joulin, Ishan Misra
May, 2021 In CVPR, 2021
Self-Supervised Pretraining of 3D Features on any Point-Cloud
Self-Supervised Pretraining of 3D Features on any Point-Cloud

SOTA 3D detection/segmentation results by learning contrastive representations on 3D data
PDF Cite Code
Eltayeb Ahmed, Anton Bakhtin, Laurens Van Der Maaten, Rohit Girdhar
February, 2021 In ICML Workshops, 2021
Physical Reasoning Using Dynamics Aware Embeddings
Physical Reasoning Using Dynamics Aware Embeddings

Self-supervised representations for physical reasoning.
PDF Cite Code
Rohit Girdhar, Laura Gustafson, Aaron Adcock, Laurens Van Der Maaten
June, 2020 In ICML Workshops, 2021
Forward Prediction for Physical Reasoning
Forward Prediction for Physical Reasoning

Forward prediction for PHYRE benchmark.
PDF Cite Code
Rohit Girdhar, Deva Ramanan
October, 2019 In ICLR, 2020 (Oral Presentation)
CATER: A diagnostic dataset for Compositional Actions and TEmporal Reasoning
CATER: A diagnostic dataset for Compositional Actions and TEmporal Reasoning

A dataset to evaluate temporal reasoning in video models.
PDF Cite Slides Video Code
Jessica Lee, Deva Ramanan, Rohit Girdhar
October, 2019 In ICLR, 2020
MetaPix: Few-Shot Video Retargeting
MetaPix: Few-Shot Video Retargeting

A dataset to evaluate temporal reasoning in video models.
PDF Cite Slides Video Code
Rohit Girdhar, Du Tran, Lorenzo Torresani, Deva Ramanan
January, 2019 In ICCV, 2019
DistInit: Learning Video Representations Without a Single Labeled Video
DistInit: Learning Video Representations Without a Single Labeled Video

Distilling representations from image models to video models.
PDF Cite
Rohit Girdhar, João Carreira, Carl Doersch, Andrew Zisserman
December, 2018 In CVPR, 2019 (Oral Presentation)
Video Action Transformer Network
Video Action Transformer Network

Among the first applications of Transformers to model videos. SOTA results: close 2nd at AVA Challenge, CVPR'18.
PDF Cite Video
Rohit Girdhar, Georgia Gkioxari, Lorenzo Torresani, Manohar Paluri, Du Tran
December, 2017 In CVPR, 2018
Detect-and-Track: Efficient Pose Estimation in Videos
Detect-and-Track: Efficient Pose Estimation in Videos

Human keypoint tracking approach that ranked first in ICCV 2017 PoseTrack keypoint tracking challenge!
PDF Cite Code
Rohit Girdhar, Deva Ramanan
November, 2017 In NeurIPS, 2017
Attentional Pooling for Action Recognition
Attentional Pooling for Action Recognition

Among the first applications of attention for contemporary video/action understanding.
PDF Cite Code
Rohit Girdhar, Deva Ramanan, Abhinav Gupta, Josef Sivic, Bryan Russell
April, 2017 In CVPR, 2017
ActionVLAD: Learning spatio-temporal aggregation for action classification
ActionVLAD: Learning spatio-temporal aggregation for action classification

Aggregating visual features for action recognition.
PDF Cite Video Code
Xiaolong Wang, Rohit Girdhar, Abhinav Gupta
March, 2016 In CVPR, 2017 (Spotlight Presentation)
Binge Watching: Scaling Affordance Learning from Sitcoms
Binge Watching: Scaling Affordance Learning from Sitcoms

Learning how humans interact with their environment by watching TV.
PDF Cite
Rohit Girdhar, David F. Fouhey, Mikel Rodriguez, Abhinav Gupta
March, 2016 In ECCV, 2016 (Spotlight Presentation)
Learning a Predictable and Generative Vector Representation for Objects
Learning a Predictable and Generative Vector Representation for Objects

A single embedding space, good for both generating and understanding 3D models
PDF Cite Video Code