This repository concludes our final project for the ”Deep Learning” course. We present a model designed to recognize sports players and their actions on the court, incorporating various training methods and pre-trained models. We achieved a 60% recognition accuracy and an inference time that qualifies as real-time recognition. We enjoyed implementing the diverse techniques we acquired throughout the semester and creating a project that combines both our academic learning and personal interests.

•⁠ ⁠Installation •⁠ ⁠Usage •⁠ ⁠Contributing •⁠ ⁠License

To install our project, first clone the code to your machine:

git clone https://github.com/your-username/your-repository.git

Then, create an enviorment for the project. A quick set-up can be accessed using the enviorment.yml and conda:

conda env create -f environment.yml

To run the model on your own video, use:

python inference.py --video <path_to_your_video>

Please notice the model only exepcts .mp4 videos. The tagged video will be saved to the output folder. You can also run the model on one of our examples:

python main.py --video "videos/Knicks3pointer.mp4"

If you want to train the model again, you first need to download the data set from link. Then, you can run the following command:

python main.py --train <output_path>

We'll be happy to answer questions and provide further information on our academic emails! {yarin.bekor,tal.dugma,yonatan.a}@campus.technion.ac.il

This project is licensed under the GPL lisence.