Drowning Detection System

A cleaned up snapshot of my LRCN (Long-term Recurrent Convolutional Network) based drowning detection project. The repository now bundles the original reports, demo photos and research notebook in a tidy structure, plus a helper script for painting the initials KA on a GitHub contribution graph.

Repository layout

• notebooks/drowning_detection_lrcn.ipynb – end-to-end experiments for training and evaluating the LRCN model on pool surveillance footage.
• docs/ – the supporting research reports (DOCX/PDF) that explain the problem framing, dataset collection and model design choices.
• media/reference_photos/ – reference stills from the recorded drowning simulations, renamed sequentially for easier browsing.
• scripts/ka_commit_art.py – utility to generate backdated commits that draw "KA" on the GitHub contribution heatmap.

Getting started

1. Create a virtual environment (Python 3.9+ recommended) and install the dependencies you need for running the notebook (TensorFlow / PyTorch, OpenCV, NumPy, etc.). The exact framework mix depends on how you want to re-run the experiments, so use the versions that work best on your machine.
2. Launch JupyterLab or VS Code, open notebooks/drowning_detection_lrcn.ipynb, and step through the workflow: preprocessing → sequence generation → LRCN training → evaluation.
3. The docs/ folder contains narratives and diagrams that complement the notebook.