Pose Aligned Networks for Deep Attribute Modeling matlab code used for the publication here: http://arxiv.org/abs/1311.5591
Switch branches/tags
Nothing to show
Clone or download
Ning Zhang
Latest commit c886076 Mar 27, 2015
Permalink
Failed to load latest commit information.
matlab Ning add missing file Mar 27, 2015
poselet_models Ning remove weights fiels Dec 22, 2014
CONTRIBUTING.md Update CONTRIBUTING.md May 12, 2014
LICENSE Update LICENSE May 12, 2014
PATENTS Update PATENTS May 12, 2014
README.md Update README Dec 23, 2014

README.md

PANDA code

PANDA (Pose-Aligned Networks for Deep Attibute) is an attribute classification system using deep neural nets. This work is created by Ning Zhang, Manohar Paluri, Marc'Aurelio Ranzato, Trevor Darrell and Lubomir Bourdev.

Citing this work

If you are using this code for your research, please cite the following paper:

@inproceedings{ZhangCVPR14,
    Author = {Zhang, Ning and Paluri, Manohar and Rantazo, Marc'Aurelio and Darrell, Trevor and Bourdev, Lubomir},
    Title = {PANDA: Pose Aligned Networks for Deep Attribute Modeling},
    Booktitle = {Conference on Computer Vision and Pattern Recognition (CVPR)},
    Year = {2014}
}

Prerequisites

  1. Caffe
  1. Poselet Detection (only if you need to run on your own data)

We provide cached poselet detections for Berkeley Attribute Dataset used in the papaer. If you want to use your own data, you have to

  • Download poselet detection code from http://www.cs.berkeley.edu/~lbourdev/poselets/
  • Run poselet detection on your images. The code returns poselet activations clustered into person hypotheses
  • Match the hypotheses to your ground truths, taking into account the overlap and score. See Section 5 of [1] for details.
  • Substract 1 from poselet_id inside phits to make them compatible to the PANDA version.
  1. MATLAB
  • The software is tested on MATLAB R2012b and R2012a versions.

Usage

License

This software is under BSD License, please refer to LICENSE file. We also provide an additional patent grant.

Bug report

If you have any issues running the codes, please report issues on github page. If you want to contribute to the codes, please follow the instructions in CONTRIBUTING.md. If you have any questions about the paper, please contact Ning Zhang (nzhang@eecs.berkeley.edu).

[1] Lubomir Bourdev and Jitendra Malik. Poselets: Body Part Detectors Trained Using 3D Human Pose Annotations. In ICCV 2009.