Files
fast-depth-tf/losses.py
Piv d88e9d3f12 Add dense-depth and experimental dense-net-nnconv5 models
Since dense-depth will use half labels by default, the nyu train/eval datasets can be loaded from here at half resolutions for labels
2021-03-29 17:59:18 +10:30

689 B