Remove speculative band labels; revert class names to English

Translate UI to French; keep code comments in English

Fix Gradio 6 compat: move css to launch(), add server_name for Spaces

Redesign: five-step pedagogical flow with spectral baseline

Fix mask interpretation: treat value 0 as background, not Water

Fix spatial size mismatch for patch sizes larger than image dimensions

Fix U-Net skip connection size mismatch for non-power-of-8 patch sizes

Add per-class pixel distribution to diagnostics, raise patch slider to 512

Fix patch extraction to always sample image edges

Robust mask remapping: auto-detect class values, add diagnostics, fix shape mismatch

Fix mask filenames: TRAINING.tif / GROUND TRUTH.tif

Switch to real dataset: 4-class, rasterio patch loader, ignore unlabeled pixels

Move pixel info above fold, bind click to both rgb+overlay, add error surfacing

Simplify compare to 2 cols, fix pixel click by keeping columns always in DOM

Fix: add training progress bar, dynamic column visibility, pixel click on RGB

Refactor: replace fixed A/B/C slots with named model cache and dropdown comparison

Upload 7 files

Delete visualize.py

Delete train.py

Delete model.py

Delete metrics.py

Delete data.py

Delete config.py

Delete app.py

Upload app.py

Delete app.py

Upload 6 files

Upload app.py

Delete app.py

Update requirements.txt

Upload 2 files

initial commit