Bias Eliminate Domain Adaptive Pedestrian Re-identification [Technique Report]
This repo contains our code for VisDA2020 challenge at ECCV workshop.
Introduction
This work mainly solve the domain adaptive pedestrian re-identification problem by eliminishing the bias from inter-domain gap and intra-domain camera difference.
This project is mainly based on reid-strong-baseline.
Get Started
- Clone the repo
git clone https://github.com/vimar-gu/Bias-Eliminate-DA-ReID.git
- Install dependencies:
- pytorch >= 1.0.0
- python >= 3.5
- torchvision
- yacs
- Prepare dataset. It can be obtained from Simon4Yan/VisDA2020.
- We use ResNet-ibn and HRNet as backbones. ImageNet pretrained models can be downloaded in here and here.
Run
If you want to reproduce our results, please refer to [VisDA.md]
Results
The performance on VisDA2020 validation dataset
| Method |
mAP |
Rank-1 |
Rank-5 |
Rank-10 |
| Basline |
30.7 |
59.7 |
77.5 |
83.3 |
| + Domain Adaptation |
44.9 |
75.3 |
86.7 |
91.0 |
| + Finetuning |
48.6 |
79.8 |
88.3 |
91.5 |
| + Post Processing |
70.9 |
86.5 |
92.8 |
94.4 |
Trained models
The models can be downloaded from:
The camera models can be downloaded from:
Some tips
- By our experience, there can be a large fluctuation of validation scores which are not completely positive correlated to the scores on testing set.
- We have fixed the random seed in the updates. But there might still be some difference due to environment.
- Multiple camera models in the testing phase may boost the performance by a little bit.