The Visual Learning Group researches methods to learn models of the real-world from images and video. Our most recent work leverages the framework of deep learning to address challenging problems at the boundary between computer vision and machine learning. Projects include image categorization, action recognition, depth estimation from single photo, as well as 3D reconstruction of human movement from monocular video.



  1. MaskProp nominated for CVPR 2020 paper award and ranked first in one of the EPIC-Kitchens CVPR 2020 challenges:

  2. Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation, by Gedas Bertasius and Lorenzo Torresani, was one of the 29 paper award nominees at CVPR 2020 (out of 1470 accepted papers).

  3. Our software submission based on this paper received the First Place Award on Unseen Kitchens and the Third Place Award on Seen Kitchens at the IEEE CVPR 2020 EPIC-Kitchens Object Detection in Video Challenge.

  1. New papers on arXiv:

  2. Video Understanding as Machine Translation,
    with Bruno Korbar, Fabio Petroni, and Rohit Girdhar.

  3. Self-Supervised Learning by Cross-Modal Audio-Video Clustering,
    with Humam Alwassel, Dhruv Mahajan, Bernard Ghanem, and Du Tran.

  4. UniDual: A Unified Model for Image and Video Understanding,
    with Yufei Wang, and Du Tran.

  5. Only Time Can Tell: Discovering Temporal Data for Temporal Modeling,
    with Laura Sevilla-Lara, Shengxin Zha, Zhicheng Yan, Vedanuj Goswami, and Matt Feiszli.

  6. Attentive Action and Context Factorization,
    with Yang Wang, Vinh Tran, Gedas Bertasius, and Minh Hoai.

  1. Three new papers to be presented at CVPR 2020:

  2. Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation,
    with Gedas Bertasius. Oral presentation.

  3. Listen to Look: Action Recognition by Previewing Audio,
    with Ruohan Gao, Tae-Hyun Oh, and Kristen Grauman.

  4. Correlation Networks for Video Classification,
    with Heng Wang, Du Tran, and Matt Feiszli.

  1. New paper to be presented at AISTATS 2020:

  2. Stein Variational Inference for Discrete Distributions,
    with Jun Han, Fan Ding, Xianglong Liu, Jian Peng, and Qiang Liu.

  1. Two new papers presented at NeurIPS 2019:

  2. Learning Temporal Pose Estimation from Sparsely-Labeled Videos,
    with Gedas Bertasius, Christoph Feichtenhofer, Du Tran, Jianbo Shi.

  3. STAR-Caps: Capsule Networks with Straight-Through Attentive Routing,
    with Karim Ahmed.

  1. Four new papers presented at ICCV 2019:

  2. SCSampler: Sampling Salient Clips from Video for Efficient Action Recognition,
    with Bruno Korbar, and Du Tran. Oral presentation.

  3. Video Classification with Channel-Separated Convolutional Networks,
    with Du Tran, Heng Wang, and Matt Feiszli.

  4. DistInit: Learning Video Representations without a Single Labeled Video,
    with Rohit Girdhar, Du Tran, and Deva Ramanan.

  5. HACS: Human Action Clips and Segments Dataset for Recognition and Temporal Localization,
    with Hang Zhao, Zhicheng Yan, and Antonio Torralba.

  1. We released a new dataset for action recognition and temporal localization, named HACS. It includes over 1.5M video clips. Give it a try!

  1. Paper presented at NIPS 2018:

  2. Cooperative Learning of Audio and Video Models from Self-Supervised Synchronization,
    with Bruno Korbar, and Du Tran.

  1. Three papers presented at ECCV 2018:

  2. Object Detection in Video with Spatiotemporal Sampling Networks,
    with Gedas Bertasius, and Jianbo Shi.

  3. MaskConnect: Connectivity Learning by Gradient Descent,
    with Karim Ahmed.

  4. Scenes-Objects-Actions: A Multi-Task, Multi-Label Video Dataset,
    with Jamie Ray, Heng Wang, Du Tran, Yufei Wang, Matt Feiszli, and Manohar Paluri.

  1. Paper presented at BMVC 2018:

  2. Self-Supervised Feature Learning for Semantic Segmentation of Overhead Imagery,
    with Suriya Singh, Anil Batra, Guan Pang, Saikat Basu, Manohar Paluri, and C.V. Jawahar.

  1. Three papers on video models presented at CVPR 2018:

  2. A Closer Look at Spatiotemporal Convolutions for Action Recognition,
    with Du Tran, Heng Wang, Jamie Ray, Yann LeCun, and Manohar Paluri.

  3. Detect-and-Track: Efficient Pose Estimation in Videos,
    with Rohit Girdhar, Georgia Gkioxari, Manohar Paluri, and Du Tran.

  4. What Makes a Video a Video: Analyzing Temporal Information in Video Understanding Models and Datasets,
    with De-An Huang, Vignesh Ramanathan, Dhruv Mahajan, Juan Carlos Niebles,
    Fei-Fei  Li, and Manohar Paluri.

  1. Together with collaborators, we organized two workshops at CVPR 2018:

  2. Brave New Ideas for Video Understanding

  3. DeepGlobe: A Challenge for Parsing the Earth through Satellite Images

  1. We released our new dataset for video comprehension, named VideoMCC. It includes over 600 hours of video. Give it a try!


We thank the following sources for supporting our research.

Logo design by Christine Claudino