Skip to content

Latest commit

 

History

History
41 lines (21 loc) · 1.38 KB

README.md

File metadata and controls

41 lines (21 loc) · 1.38 KB
Raw

Dynamic Feature Pruning and Consolidation for Occluded Person Re-Identification

The official repository for Dynamic Feature Pruning and Consolidation for Occluded Person Re-Identification achieves state-of-the-art performances on occluded person Re-ID.

News

  • 2023.03 FPC's code will be released coming soon.
  • 2023.11We first updated the evaluation code, and we will continue updating.

Pipeline

framework

Visualization of the Patch Pruning Process

framework

Evaluation

1. Requirements Installation

Install the necessary packages using the provided env.yaml file:

2. Prepare Datasets

Download the Occluded or Holistic Person ReID datasets (e.g., Occluded-Duke). Unzip them and rename the folders under the ./data directory.

3. Prepare Checkpoints

Download Checkpoints from this link. Save the checkpoint file as ./FPC_Occ_Duke_reconstruction.pth.

4. Evaluate

Run the evaluation using the following command: python test.py --config_file configs/OCC_Duke/vit_transreid_stride.yml MODEL.DEVICE_ID "('0')" TEST.WEIGHT './FPC_Occ_Duke_reconstruction.pth'

Acknowledgements

Our code is heavily built on TransReid.