Home

Awesome

Tracking Persons-of-Interest via Adaptive Discriminative Features (ECCV 2016)

This is the research code for the paper:

Shun Zhang, Yihong Gong, Jia-Bin Huang, Jongwoo Lim, Jinjun Wang, Narendra Ahuja and Ming-Hsuan Yang. "Tracking Persons-of-Interest via Adaptive Discriminative Features", in Proceedings of European Conference on Computer Vision (ECCV), 2016.

We take the T-ara sequence as an example to evaluate our adaptive feature learning approach in this code. Our project website can be found here:

Project page

Citation

If you find the code and pre-trained models useful in your research, please consider citing:

 @inproceedings{Zhang-ECCV-2016,
       author = {Zhang, Shun and Gong, Yihong and Huang, Jia-Bin and Lim, Jongwoo and Wang, Jinjun and Ahuja, Narendra and Yang, Ming-Hsuan},
       title = {Tracking Persons-of-Interest via Adaptive Discriminative Features},
       booktitle = {European Conference on Computer Vision},
       year = {2016},
       pages = {415-433}
   }

System Requirements

Installation

  1. Download and unzip the project code.

  2. Install Caffe. Please follow the Caffe installation instructions to install dependencies and then compile Caffe:

    # We call the root directory of the project code `AFL_ROOT`.
cd $AFL_ROOT/external/caffe-Triplet-New
make all -j8
make pycaffe
make matcaffe

  3. Download the T-ara images and extract all images into AFL/data/Tara.

  4. Download the AlexNet model:

    cd $AFL_ROOT/external/caffe-Triplet-New
scripts/download_model_binary.py models/bvlc_reference_caffenet

  5. Download the VGG-Face Model and put it in $AFL_ROOT/external/caffe-Triplet-New/models/VGG. Download the pre-trained face model and put it in $AFL_ROOT/external/caffe-Triplet-New/models/pretrained_web_face.

Usage

Directly run the script run_Tara_example.sh.

Or run the following commands step by step:

  1. Mine constraints:

    cd $AFL_ROOT
# Start MATLAB
matlab
>> genTracklet('Tara')

  2. Learn adaptive discriminative features:

    cd $AFL_ROOT
sh shell_scripts/Tara/adapt_Triplet.sh

  3. Extract features:

    sh shell_scripts/Tara/extract_All_Feas.sh

  4. Perform hierarchical agglomerative clustering algorithm (you can get Fig. 6(a) in our supplementary materials):

    matlab
>> clustering_tracklets('Tara')

  5. Perform a simple multi-face tracking:

    matlab
>> facetracking('Tara')