Add tiling support for all newly integrated models that currently lack it, including the latest DEIMV2, RFDETR and YOLO26 architectures. This work includes creating and validating tiling training recipes, ensuring compatibility with tiled datasets, and verifying end-to-end functionality of automatic tiling workflows.
Goals
- Add tiling recipe files for all newly supported models
- Enable training on tiled datasets
- Ensure compatibility with the Auto Tiling feature
- Validate training, validation, and inference using tiled datasets
- Verify consistent behavior across supported tasks
Acceptance Criteria
- All newly added models support tiling through dedicated recipe files
- Models can be trained and evaluated on tiled datasets without additional configuration
- Auto Tiling works correctly with all supported models
- End-to-end tests validate the complete tiling workflow
- Documentation is updated with any model-specific tiling requirements or limitations
Add tiling support for all newly integrated models that currently lack it, including the latest DEIMV2, RFDETR and YOLO26 architectures. This work includes creating and validating tiling training recipes, ensuring compatibility with tiled datasets, and verifying end-to-end functionality of automatic tiling workflows.
Goals
Acceptance Criteria