This repository contains the implementation of paper "Pose-Aware Person Recognition" by Vijay Kumar, Anoop Namboodiri, Manohar Paluri, C V Jawahar published at CVPR17.

The implementation is based on Python Caffe.

Datasets:

1. Download the datasets from the below links and place in data/ folder.
2. PIPA (test): Link
3. Hannah movie : Link
4. IMDB : Link
5. Soccer videos : Link

Models:

1. Download the trained models and place in models/ folder.
2. The models (baseline, pose-specific and pose estimator) are available at link

Testing:

Dependencies: Liblinear.

1. To reproduce the results on PIPA test set, run run_PIPA.ipynb
2. For recognition in movie scenario, run run_hannah.ipynb
3. For recognition in soccer setting, run run_soccer.ipynb
4. Change the data folder variable in these scripts according to your path.
5. Replace the liblinear path to your correct liblinear installation directory.

References:

If you use this code or data, please cite the following papers.

1. Vijay Kumar, Anoop Namboodiri, Manohar Paluri, C V Jawahar, Pose-Aware Person Recognition, CVPR 2017.
2. N. Zhang et al., Beyond Fronta Faces: Improving Person Recognition using Multiple Cues, CVPR 2014.
3. Oh et al., Person Recognition in Personal Photo Collections, ICCV 2015.
4. Li et al., A Multi-lvel Contextual Model for Person Recognition in Photo Albums, CVPR 2016.
5. Ozerov et al., On Evaluating Face Tracks in Movies, ICIP 2013.