Mask-Net

Mask-Net is a U-Net-based architecture designed for cloud masking in Sea Surface Temperature (SST) imagery.

The model combines:

• Spatial Attention
• Channel Attention
• Inter-Level Fusion Module (ILFM)

to improve cloud-ocean discrimination in dynamic coastal upwelling regions.

Architecture

Mask-Net extends the classical U-Net architecture by introducing an Inter-Level Fusion Module (ILFM) within the decoder skip connections.

The ILFM combines:

• Spatial attention
• Channel attention

through additive fusion followed by convolutional refinement.

Requirements

pip install -r requirements.txt

Quick Test

python models/masknet.py

Expected output:

Input : torch.Size([1, 1, 288, 288])
Output: torch.Size([1, 1, 288, 288])

Citation

If you use this code, please cite:

@article{MASKNET2026,
  title={Mask-Net: Cloud Masking for Sea Surface Temperature Imagery Using Spatial and Channel Attention},
  author={Kouassi Adelphe Christian N’GORAN, XXX},
  year={2026}
}