TODO.md

TODO

Planned work for this repository.

1. Complete evaluation of the current best checkpoint on the NYU Depth V2 Eigen test split and report the standard depth-estimation metrics.
2. Continue training until the validation metric plateaus, then select the final checkpoint.
3. Extend the evaluation protocol with an outdoor benchmark (Virtual KITTI 2) to track generalization beyond the indoor training distribution.
4. Finalize the model card with the deployed configuration and benchmark numbers.
5. Release the final adapter weights.