Seafloor sediment segmentation (U-Net)
Semantic segmentation of seafloor imagery + bathymetry across 17 sediment classes, U-Net with ResNet backbones.
- When
- 2025-02 – 2025-07
- Result
- IoU ~0.74 on held-out set · 17 substrate classes
- Python
- PyTorch
- torchvision
- OpenCV
- GeoPandas
- scikit-learn
Research work from the UD DeepREAL Lab. RGB imagery paired with bathymetry, class-weighted cross-entropy, cosine-annealed LR, IoU/Dice evaluation. Paired with the FathomNet scaling study as one marine-CV story.