Datasets, Codes, Videos


  • IEEE TPAMI'19 (submission) + ECCV'18, Privacy attributes in HMDB51 (PA-HMDB51) dataset [Website]

    • the first and currently largest benchmark for privacy-preserving video action recognition. Released in June 2019.

  • CVPR'19, Multi-Purpose Image Deraining (MPID) dataset [Website]

    • the first and currently largest benchmark for image deraining research. Released in March 2019.

  • IEEE TIP'19, REalistic Single Image DEhazing (RESIDE) dataset [Website]

    • the first and currently largest benchmark for image dehazing research. Released in December 2017.

  • ACM MM'15, Adobe Visual Font Recognition (VFR) dataset [Link]

    • the first large-scale, fine-grained benchmark of font text images, for the task of font recognition and retrieval. Released in May 2015.

Paper Codes

We embrace open-source! Our recent codes can be found at VITA Group GitHub Repository:

Some legacy code repos (that are elsewhere hosted):

  • IEEE TCSVT'19, Frank-Wolfe network for image restoration [GitHub]

  • ICML'19, Plug-and-play optimization with deep learning [GitHub]

  • Bioinformatics'19, DeepAffinity model for protein-compound affinity prediction [GitHub]

  • IEEE TMI'19, Convolutional sparse coding for compressed sensing CT [GitHub]

  • ICASSP'19, GAN-based data augmentation [GitHub]

  • KDD'18, Subspace deep network [GitHub]

  • IJCAI'18, Segmentation-aware denoising [GitHub]

  • ICCV'17, AOD-Net for dehazing [Official Caffe GitHub] [Keras] [Pytorch]

  • IEEE TIP'17, Greedy batch-based minimum-cost flow tracking [GitHub]

  • ACII'17, Low-resolution facial expression recognition [GitHub]

  • BMVC'17, Transformed anti-sparse hashing [GitHub]

Other Software

  • DeepFont, an Adobe Photoshop built-in feature, as one of its leading contributors [Official Photoshop manual]

  • Distributed Machine Learning (DML) ToolBox, as part of Microsoft Prajna.

  • Dehazing algorithms integrated in the Computer Vision Emergency Response (CVER) toolkit [by students Pritish Uplavikar and Ryan Wells]

Selected Talk Records

  • Georgia Tech Machine Learning Center Seminar (2019) "Learning to Optimize from Data: Faster, Better, and Guaranteed" [Video]

  • Summer School, at University of Sao Paulo, Brazil (2018): Deep learning short course [Video]; and invited talk on dehazing [Video]

  • CVPR (2017), Tutorial on "Dealing with Reality: Low-Quality Visual Data Processing and Analytics" [Video]

  • DeepFont (ACM MM'15) talks: Adobe MAX conference 2015 demo, GTC 2016 tutorial, among many others.