Skip to content

Laboratories for media processing with ML for Sport events

Notifications You must be signed in to change notification settings

rodzanto/ml-sports-media

Repository files navigation

ml-sports-media

Laboratories for automatic sport's media processing with AWS AI & ML services.

Pre-requisites and notes:

  1. Each of the following labs include a Jupyter notebook for following the instructions in the cells and code; the notebooks are designed to be deployed with Amazon SageMaker. You can either use SageMaker Studio IDE or a traditional SageMaker Notebook instance. For instructions on how to onboard with SageMaker Studio visit: https://docs.aws.amazon.com/sagemaker/latest/dg/gs-studio-onboard.html

  2. For starting the work with the labs first clone this repository to your notebook environment. In SageMaker Studio go to "File", "New", "Terminal", then in the new terminal window write:

git clone https://github.com/rodzanto/ml-sports-media

Studio

  1. Go to the file browser in the left pane, you should now see the folder with the copy of the "ml-sports-media" repository. You can access the Lab #1 and open the corresponding Jupyter notebook. Follow the instructions in the notebook.

  2. Proceed with Lab #2.

  3. Proceed with Lab #3.

Goal: Classify and track objects (e.g. ball, specific players) in stored videos for sports (e.g. football, basketball). Rely on Amazon SageMaker for this (Ground Truth, Training, Endpoints).

Lab1_1

Lab1_2

Lab1_2_1

Goal: Classify activities (e.g. shot, pass, nice play) in stored videos for sports (e.g. football, basketball). Rely on Amazon SageMaker for this (Ground Truth, Training, Endpoints).

Lab2

Goal: Process videos of sports based on activity classification for automatically clipping highlights with relevant plays. Rely on AWS AI Services for this (Amazon Transcribe, Amazon Textract, Amazon Rekognition, etc.).

Lab2

About

Laboratories for media processing with ML for Sport events

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published