HYPSO-CNN v1.0.0 – Initial Release

This is the first official release of HYPSO-CNN, a PyTorch-based 1D-CNN implementation for sea/land/cloud classification of hyperspectral satellite images from the NTNU SmallSat Lab HYPSO-2 mission.

Included

• Full implementation of 1D-JustoLiuNet, based on Justo et al. (2023)
• Pipeline following the SoC (Separation of Concerns) principle
• Functional decomposition through clearly scoped utility functions
• Layered architecture organized into data, domain, application, and presentation layers
• Optimized for NVIDIA GPU clustering and CUDA 12.6 compatibility
• A complete final report PDF with theoretical grounding, results, and discussion

Notes
This release includes observations from internal testing related to model robustness (e.g., sensitivity to noise and occasional label inconsistencies). These findings provide valuable feedback for future refinement and are discussed in the accompanying report.

How to Use
See the README for full setup, usage, and diagrams.