leonel11 / DetectTrackSportEvents

Program for detection and tracking players on a sports ground and calculation of basic statistical indicators using deep learning

Sport AI System

Program for analysis of videos of sport events based on machine learning

Prerequisites

• OS: Linux (tested only on Ubuntu >= 16.04) or Windows 10
• NVIDIA videocard with CUDA capability >= 3.5
• Installed CUDA >= 9.0 and cuDNN 7

Requirements

• Towards-Realtime-MOT
• Python >= 3.6
• PyTorch >= 1.2.0
• opencv-python
• motmetrics
• NumPy
• logging
• Cython
• cython-bbox
• FFmpeg
• numba
• matplotlib
• sklearn
• Pillow 6.1.0
• tqdm
• pandas
• yagmail
• SciPy
• argparse
• PyQt5
• plotly
• your favourite browser

Installation

1. Clone this repository

   git clone https://github.com/leonel11/DetectTrackSportEvents.git

2. Clone this repository into another directory

   git clone https://github.com/Zhongdao/Towards-Realtime-MOT

   or download it as a zip file and repack

3. Copy all files of repository Towards-Realtime-MOT to folder video_player without exchanging files of the same name

4. Exchange file video_player/models.py to file of the same name from folder video_player/exchange_files/ with the replacement

5. Install all requirements (you can follow some instructions of installation using Requirements or Issues in case of any problem)

6. Copy file of weights JDE-1088x608 (1 or 2) for running of MOT algorithm

Advice

1. For Ubuntu:

   • before start of installation type this command:

     sudo apt-get update

   • After logging installation, before Cython installation type this command

     sudo apt install libsm6 libxrender-dev

2. For Windows:

   FFmpeg installation

3. It's possible to work with virtualenv environment of Python. You can create it after Python installation, before PyTorch installation. Also read this article which describes how to work with virtualenv.

Docker

It's also possible to launch this GUI application using Docker container.

1. Install Docker on your computer

2. Pull and run container with the support of CUDA >=9.0 and cuDNN 7. For example, 1, 2 etc.

3. After PyQt5 installation, before plotly installation type these commands into container:

   sudo apt-get update
   
   export QT_DEBUG_PLUGUINS=1
   
   sudo apt-get install libxcb-randr0-dev libxcb-xtest0-dev libxcb-xinerama0-dev libxcb-shape0-dev libxcb-xkb-dev
   
   sudo apt install libxkbcommon-x11-0

4. Install your favourite browser into container

5. For Windows:

   • Use XLaunch to implement your own server. Here is some instructions how to launch it.
   • For volume project don't forget to share needed drive in Docker settings

Running

• For Windows: double click on video_player/SportAISystem.lnk or run video_player/run_sportaisystem.bat in cmd
• For Linux or Docker container: run video_player/sportaisystem.sh in Terminal