This page list the codes, models, and softwares we released in the last years. Most of them are developed and maintained by our students and collaborators. The topics includes the following areas:
  • Action Recognition and Detection in Videos
  • Scene Understanding and Classification
  • Scene Text Detection and Recognition
  • Face Detection and Recognition
  • Action Recognition and Detection in Videos

  • Temporal Segment Networks for action video classification(ECCV 16 paper, 1st in ActivityNet Challenge 2016) [Codes]
  • Enhanced Motion Vector CNNs for fast Action Recognition (CVPR 16 paper, realtime action recognition, 390 frames per second) [Project] [Codes]
  • Trajectory-Pooled Deep-Convolutional Descriptors for Action Recognition(CVPR 15 paper, high Performance on HMDB51 (65.9%) and UCF101(91.5%).) [Project] [Codes]
  • Multi-View Super Vector for Action Recognition(CVPR 14 Oral paper, a new method for jointly encoding multiple types of local descriptors.) [Code]

    Scene Understanding and Classification

  • MR-CNNS for large scale scene classification (2nd in scene classification task ImageNet 2016, 1st in LSUN 2016) [Code]
  • Weakly Supervised PatchNets for Scene Recognition ( MIT Indoor67 (86.2%) and SUN397 (73.0%) )[Paper] [Code]
  • Pairwise Rotation Invariant Co-occurrence Local Binary Pattern ( A very fast local descriptors for scene and texture classification )[PAMI 14 Paper][Project] [Code]
  • Scene Text Detection and Recognition

  • Connectionist Text Proposal Network for Scene Text Detection[ECCV 16 Paper] [Online demo]
  • Facial Detection and Recognition

  • A Discriminative Deep Feature Learning Approach for Face Recognition(Single model trained on CAISA-WebFace achieves ~99% verification rate on LFW) [ECCV 16 Paper] [Code] [Model]
  • Multi-task Cascaded Convolutional Neural Networks for Joint Face Detection and Alignment [SPL 16 Paper] [Code]
  • Back to Top