diff options
| author | skal <pascal.massimino@gmail.com> | 2026-02-13 17:46:09 +0100 |
|---|---|---|
| committer | skal <pascal.massimino@gmail.com> | 2026-02-13 17:46:09 +0100 |
| commit | a7340d378909cadbfd72dbd1f5b756f907c2a3e0 (patch) | |
| tree | 60a34dd084d746f0c6dad50b0d5cc7f20bc0c409 /tools/gen_test_tga.cc | |
| parent | f6b3ea72a03850654b69986bc82bb249aaabe2e3 (diff) | |
CNN v2 training: Add --grayscale-loss option for luminance-based loss computation
Add option to compute loss on grayscale (Y = 0.299*R + 0.587*G + 0.114*B) instead of full RGBA channels. Useful for training models that prioritize luminance accuracy over color accuracy.
Changes:
- training/train_cnn_v2.py: Add --grayscale-loss flag and grayscale conversion in loss computation
- scripts/train_cnn_v2_full.sh: Add --grayscale-loss parameter support
- doc/CNN_V2.md: Document grayscale loss in training configuration and checkpoint format
- doc/HOWTO.md: Add usage examples for --grayscale-loss flag
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Diffstat (limited to 'tools/gen_test_tga.cc')
0 files changed, 0 insertions, 0 deletions