This is the code for the drowning detection system proposed in the report.
To install the required libraries do:
pip install -r requirements.txt
The code for the integrated drowning detection system can be found in the 'code' directory. There are 4 versions of the final system, each employing different model:
integrated_baseline.py,
integrated_cnnsingle.py,
integrated_cnnseq.py and
integrated_cnnrnn.py.
To run the system on a specified footage: python integrated_{model_name}.py {PATH_TO_FOOTAGE}
Other than those scripts, there are the following directories:
models: Saved models and their training scripts.testing: Test scripts used to evaluate models and results.hp_tuning: Hyperparameter tuning scripts and results.datasets: Scripts used to create PyTorch datasets and keypoints.system_tests: Old integrated system scripts used to test the latency and jitter.utils: Scripts used for data handling and keypoint extraction with YOLOv7.
What is not included in this repository:
- YOLOv7 Scripts and models.
- The "Water Behavior" dataset.