Marine Waste Detection is an AI-based project designed to automatically classify and segment marine debris from underwater sonar images.
The main goal is to support ocean cleanup initiatives by detecting and analyzing pollution sources in marine environments.
The first objective focuses on classifying the detected debris based on predefined categories.
The training dataset was organized into folders labeled by class names.
Class labels used:
CLASS_NAMES = [
'bottle', 'can', 'chain', 'drink-carton', 'hook',
'propeller', 'shampoo-bottle', 'standing-bottle', 'tire', 'valve'
]📓 A training notebook is available inside the training/ directory for model reproduction and experimentation.
The second phase focuses on precise segmentation of marine debris in sonar images. The original dataset lacked segmentation masks, which are required for training segmentation models.
To overcome this limitation, we used the SAM (Segment Anything Model) to:
Automatically generate high-quality masks for each object.
Avoid manual annotation efforts.
Build a clean dataset suitable for training a segmentation network.
🧠 Both models (classification and segmentation) were trained using TensorFlow.
Category Tools / Frameworks Programming Language Python Deep Learning Framework TensorFlow Segmentation Tool SAM (Segment Anything Model) Web Deployment Flask Visualization / Notebooks Jupyter Notebook