/PersonReID-CACENET

TensorFlow implementation of our paper "Devil’s in the Details: Aligning Visual Clues for Conditional Embedding in Person Re-Identification"

Primary LanguagePythonOtherNOASSERTION

Devil’s in the Details: Aligning Visual Clues for Conditional Embedding in Person Re-Identification

Introduction

This is the tensorflow implementation of our paper "Devil’s in the Details: Aligning Visual Clues for Conditional Embedding in Person Re-Identification". In this paper:

  1. CACE-Net is able to integrate both visual clue alignment and conditional feature embedding into a unified ReID framework
  2. Instead of using a pre-defined Adjacency Matrix, our CACE-Net uses a novel correspondence attention module where the visual clues is automatically predicted and dynamically adjusted during training

image

requirements

python >= 3.6 tensorflow >= 1.8

train

Prepare Datasets

  1. File Directory: ├── partitions.pkl ├── images │ ├── 0000000_0000_000000.png │ ├── 0000001_0000_000001.png │ ├── ...

  2. Rename the images in following convention: "000000_000_000000.png" where the first substring splitted by underline is the person identity; for the second substring, the first digit is the camera id and the rest is track id; and the third substring is an image offset.

  3. "partitions.pkl" file This file contains a python dictionary storing meta data of the datasets, which contains folling key value pairs "train_im_names": [list of image names] #storing a list of names of training images "train_ids2labels":{"identity":label} #a map that maps the person identity string to a integer label "val_im_names": [list of image names] #storing a list of names of validation images "test_im_names": [list of image names] #storing a list of names of testing images "test_marks"/"val_marks": [list of 0/1] #0/1 indicates if an image is in gallery

train

  1. Configure basic settings in core/config
  2. Define the network in net and register in the factory.py
  3. Set the corresponding hyperparameters in the experiment yaml
  4. set experiment.yaml path in config.yaml
  5. python train.py

test

  1. Configure settings in eval_config.yaml， pay attetion to set train_mode false
  2. register in val.save_pb
  3. python evalution.py

example

yaml: 'experiment/graph/cacenet.yaml'

market: [mAP: 90.11%], [cmc1: 95.84%]

peformance

image

Citation

If you find this code useful, please cite the following paper:

@article{jiang2020devil,
  title={Devil's in the Detail: Graph-based Key-point Alignment and Embedding for Person Re-ID},
  author={Jiang, Xinyang and Yu, Fufu and Gong, Yifei and Zhao, Shizhen and Guo, Xiaowei and Huang, Feiyue and Zheng, Wei-Shi and Sun, Xing},
  journal={arXiv preprint arXiv:2009.05250},
  year={2020}
}