This repository provides a solution to one of the task assigned to assess the ability to solve real-world computer vision problem in sport analytics.

Identification of each player ensuring same player retains the same ID even after going out of view in a given 15-second video.

To get a local copy up and running, follow these simple steps.

Install Miniconda

Below are steps to follow in setting up the project.

1. Create a new conda environment

conda create --name pytorch python=3.10

2. Activate the environment

conda activate pytorch

3. Install dependencies

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu[CUDA_VERSION] 
pip install ultralytics
pip install streamlit
pip install ipykernel

Note: Replace [CUDA_VERSION] with the version of CUDA installed on your system. You can find this version by running the nvidia-smi command in your terminal, which displays the driver and CUDA version currently in use by your GPU. Example: If CUDA Version is 12.8, then replace [CUDA_VERSION] with 128 - pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128

4. Directory Structure

liat.ai
|   .gitignore
|   app.py
|   main.py
|   README.md
|   requirements.txt
|   Soccer Player Re-Identification Report.pdf
|   test.ipynb
|
+---assets
|       demo.gif
|
+---data
|       15sec_input_720p.mp4
|
+---models
|       best.pt
|
\---runs
    \---detect
        \---track
                15sec_input_720p.avi

5. Use the following command to run main.py

python main.py

6. Use the following command to run streamlit app

streamlit run app.py