Tennis Video Analysis

This repository basically analysis tennis video and segment it. Mian objective of this project is user will upload a video via frontend, Backend receive this video and segment and generate new segment videos based on player ball serve.

How to run

Importent first install 'ffmpeg', for linux: sudo apt install ffmpeg

1. create python virtual env. User python version 3.10

python -m venv venv

2. Activate venv [use os based command]

source venv/bin/active [Linux]

3. Install packages

pip install -r requirements.txt

4. if 'package 'lap' not found' [Optional]

pip install lap

5. Run project

python main.py

6. Default url:

localhost:8000 or 0.0.0.0:8000

But before run set .env

Create a '.env' file in project root dir. then add below variables with valid values.

AWS_ACCESS_KEY_ID=your_aws_access_key
AWS_SECRET_ACCESS_KEY=your_aws_access_kay
AWS_REGION=which_region your bucket located
IS_UPLOAD_TO_S3=true | false
IS_CLEAN_LOCAL_VIDEOS=false | true
AWS_S3_BUCKET_INPUT=your_bucket_name
AWS_S3_BUCKET_OUTPUT=your_bucket_name

IS_UPLOAD_TO_S3 - - true: cut clip would be uploaded in s3

IS_CLEAN_LOCAL_VIDEOS: - true: local saved clip will be removed.

Local saved location:

project_root:
    static:
        - uploads
        - thumbnails
        

This uploads file contain locally saved clip. I you set IS_CLEAN_LOCAL_VIDEOS = true, it will clean this folder otherwise it will keep.

Model

Download model from given link in release note and keep on bellow folders.

project_root:
    - models

Here are Example outputs:

Tennis1.mp4

Tennis2.mp4