Skip to content
/ DCSL Public

code for "Semantics-aware Deep Correspondence Structure Learning for Robust Person Re-identification"

Notifications You must be signed in to change notification settings

yaqing01/DCSL

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DCSL for Person Re-Identification (Re-ID)

Code for paper "Semantics-Aware Deep Correspondence Structure Learning for Robust Person Re-Identification"

Clone and Installation

Clone the project with Caffe submodule

$ git clone --recursive https://github.com/yaqing01/DCSL.git

Install Caffe

Download and Prepare for the Dataset

  1. Download CUHK03/01 datasets from CUHK Person Re-identification Datasets

  2. Unzip the datasets into the dataset/ folder

The image files are organized as

dataset/cuhk03/cuhk03_release/data/campair_1/01_0001_01.jpg
dataset/cuhk03/cuhk03_release/data/campair_1/01_0001_02.jpg
dataset/cuhk03/cuhk03_release/data/campair_1/01_0001_03.jpg
  1. Generate training data
cd dataset
python generate_training_pairs.py
cd ..

You only need to modify the following configurations in generate_training_pairs.py

set_no = 1 # the training/validation/test split
save_p = 'train_lmdb' # path to save lmdb
dataset_usage = [0,1,0,0,0] # dataset for evaluation

Train the network

We have written all the model templates for training in models/reid/dcsl, all we need is to generate the training protos with the specified configurations.

  1. Prepare for training
mkdir experiments
./models/reid/dcsl/prepare.sh
  1. Train the model using the generated proto files and you can download the pre-trained bvlc_googlenet.caffemode for fine-tuning.
./experiments/reid_dcsl/set01/train_model.sh [GPU-ID]
  1. Finally, you can evaluate and visualize the trained model by running code/eval/cuhk03_test.ipynb

Please cite our work in your publications if it helps your research:

@inproceedings{ZhangLZZ16,
  author = {Yaqing Zhang and Xi Li and Liming Zhao and Zhongfei Zhang},
  title = {Semantics-Aware Deep Correspondence Structure Learning for Robust Person Re-Identification}, 
  booktitle = {Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence, {IJCAI} 2016, New York, NY, USA, 9-15 July 2016},
  pages = {3545--3551},
  year = {2016},
}

About

code for "Semantics-aware Deep Correspondence Structure Learning for Robust Person Re-identification"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published