Top 2 solution for MMSports2023_-Player_Reidentification

Top 2 solution for MMSports2023_-Player_Reidentification

Requirements

• Python 3.8
• cuda 11.7
• timm 0.8.17
• pytorch 1.13.1
• torchaudio 0.13.1
• torchvision 0.14.1

About the Code

1. Prepare Data

Download the competition data

python download_data.py 

2. Creating Data Frames

python preprocess_data.py

3. Train the Model

python train.py 

4. Feature Extraction

Here the feature extraction is performed on our trained model for query and gallery.

python  fea.py 

5. Feature Enhancement

Feature enhancement is performed separately for the trained model.We use the pyretri repository（https://github.com/PyRetri/PyRetri/tree/master/pyretri） ，feature enhancement is performed on the features that have been extracted.

python post_process.py 

6. Prediction

python predict.py