- Introduced `tiles/__init__.py` to document the tile repository functionality.
- Created `tiles/build_catalog.py` for converting GeoTIFFs to Cloud-Optimized GeoTIFFs and generating a catalog in GeoJSON format.
- Added `tiles/build_samples.py` to generate map samples for the GeoLibre plugin from precomputed results.
- Implemented `tiles/catalog.py` to load and query the tile catalog based on bounding boxes.
- Developed `tiles/mosaic.py` for stitching tiles into a single scene based on a specified bounding box.
- Added `tiles/outputs.py` for writing georeferenced results, including height maps and object data in various formats.
feat: expose scaled inference API for complete Space benchmarks
Derive per-pixel height uncertainty from the adaptive-bin distribution, propagate it through tiled inference and TTA, and publish ndsm_std_m.npy for the DepthWizard frontend.
Enhance GPU processing and batching capabilities; update README for clarity on test-time augmentation and batch sizes
- Implement object extraction from segmentation and height maps in `infer/objects.py`.
- Add class map previews and detected objects overlay in `viz/classes.py`.
- Update `app.py` to save segmentation and object data as artifacts.
- Enhance README with details on object classes and 3D objects.
- Create tests for object extraction in `tests/test_objects.py`.
- Update requirements and configuration files for new dependencies.
Add synthetic sample scene imagery for the Examples widget
viewer/sample/*.png was gitignored, so the gr.Examples entry added in
app.py had no rgb.png to point at once deployed. Track the sample
imagery and regenerate the full synthetic scene via make_sample.py.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- Implemented DPT trunk architecture in models/dpt.py for feature extraction.
- Added exponential moving average functionality in models/ema.py to stabilize training.
- Created DINOv3 encoder in models/encoder.py for robust block finding and layer-wise learning rate decay.
- Developed heads for regression and classification tasks in models/heads.py, including gated fusion mechanism.
- Introduced custom loss functions in models/losses.py to address specific training challenges.
- Implemented test-time augmentation strategies in models/tta.py to enhance model predictions.
- Updated requirements.txt to include necessary dependencies for the new model components.
