https://github.com/abdo-reda/soccerglanceprototype

Currently, this is a prototype for the Action Spotting Model for the SoccerNetV2 Dataset.
https://github.com/abdo-reda/soccerglanceprototype

Last synced: 8 months ago
JSON representation

Currently, this is a prototype for the Action Spotting Model for the SoccerNetV2 Dataset.

• Host: GitHub
• URL: https://github.com/abdo-reda/soccerglanceprototype
• Owner: Abdo-reda
• Created: 2022-12-21T01:26:41.000Z (over 3 years ago)
• Default Branch: master
• Last Pushed: 2023-05-20T23:04:03.000Z (about 3 years ago)
• Last Synced: 2024-12-18T06:33:54.085Z (over 1 year ago)
• Language: Python
• Size: 30.7 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: ReadMe.md

Awesome Lists containing this project

README

          


### Setup Enviornment 

* conda create -y -n PrototypeEnv python=3.9

* conda activate PrototypeEnv

* conda install -y cudnn=8.2

* conda install -y cudatoolkit=11.3

* conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia

* pip3 install -r requirements.txt

-------------- 

### Notes & Prerequistes

* This project mainly depends that an  nginx server with an rtmp module  is already setup to accept streams on rtmp://localhost:1935/live/ 

    * However the code could be modified to accept streams from whenever, the SOURCE will just need to be modified in the server.py file.

    * Morover, The nginx.config that was used is included in the project directory for reference.

--------------

### Using the Prototype

* To activate the enviornment 

    * `conda activate PrototypeEnv`

* To run the model server 

    * `python server/server.py`

    * You can then open http://localhost:5000/ and click on process stream button to start generating processing the video stream (this will simply execute the main.py script)   

* To run an example of a client server 

    * `python client_server/client_server.py`

--------------

### Project Structure

*  Main Directory 

    * `main.py` this is the main script that is responsible for creating and executing other python scripts/processes that will process the video stream.

    * `configuration.py` is responsible for setting up parameters (a yaml file) that was used in the action spotting model (creating the features and generating actions).

    * `offline_prototype.py` this is the old script that generates actions for a downloaded video.

*  Server Directory  Contains the main logic for the model server.

    * `server.py` this is the main model server, it processes the video streams and communicates with the api.

    * `api_service.py` this is a script that contains the most common calls/functions to the api, for example sending highlights or match is live and so on ..

*  Client Server Directory 

    * `client_server.py` this an example for what a client application would look like, this was created to test that the clients can recieve the highlights.

*  Chunk_Generation 

    * `transcribe_audio.py` this script is reponsible for transcribing the audio chunks.

    * `extract_highlights.py`this script is reponsible for generating highlights from the transcribed audio chunks.

    * `extract_actions.py` this script is responsbile for generating actions from the transcriptions.

    * `capture_stream.py` this script was used to capture a video stream from interent/youtube using streamlink and store that data in temp folder.

    * `generate_chunks.py` this script was used to generate video chunks from the data that was caputred by capture_stream.py.

*  Build_Models [Archived] 

    * This contains different scripts that will reponsible for training and creating an instance of the model that was used for action spotting.

*  Feature_Generation [Archived] 

    * `generate_features.py` this script is reponsbile for extracting features which will later be used to extract actions from the video stream.

*  Output_Generation [Archived] 

    * `generate_output.py` this script is reponsible for generating actions from the extracted features from the video.