diff options
| author | skal <pascal.massimino@gmail.com> | 2026-03-27 08:41:05 +0100 |
|---|---|---|
| committer | skal <pascal.massimino@gmail.com> | 2026-03-27 08:41:05 +0100 |
| commit | 37df61d1a0dbd5e253f9db778c17c4187e453b8d (patch) | |
| tree | 2879f69f35ddeff1ead30cb5099c1073c5ab376a /src/3d/scene.h | |
| parent | fb13e67acbc7d7dd2974a456fcb134966c47cee0 (diff) | |
- Switch MSELoss → L1Loss in train_cnn_v3.py (median-seeking, avoids gray-blob)
- Normalize depth_grad channels with tanh(10x) in cnn_v3_utils.py (bounds ±∞ signed values)
- Match normalization in gbuf_pack.wgsl: tanh((right-left)*5.0) == tanh(10*central_diff)
handoff(Gemini): training pipeline only; no C++ or test changes needed.
Diffstat (limited to 'src/3d/scene.h')
0 files changed, 0 insertions, 0 deletions