diff options
| author | skal <pascal.massimino@gmail.com> | 2026-02-13 23:17:42 +0100 |
|---|---|---|
| committer | skal <pascal.massimino@gmail.com> | 2026-02-13 23:17:42 +0100 |
| commit | 6fa9ccf86b0bbefb48cefae19d4162115a3d63d3 (patch) | |
| tree | 529f68a33d9e4dcc8e473ed604c0bfb6f6f2704f /training/target_3 | |
| parent | f81a30d15e1e7db0492f45a0b9bec6aaa20ae5c2 (diff) | |
CNN v2: Alpha channel depth handling and layer visualization
Training changes:
- Changed p3 default depth from 0.0 to 1.0 (far plane semantics)
- Extract depth from target alpha channel in both datasets
- Consistent alpha-as-depth across training/validation
Test tool enhancements (cnn_test):
- Added load_depth_from_alpha() for R32Float depth texture
- Fixed bind group layout for UnfilterableFloat sampling
- Added --save-intermediates with per-channel grayscale composites
- Each layer saved as 4x wide PNG (p0-p3 stacked horizontally)
- Global layers_composite.png for vertical layer stack overview
Investigation notes:
- Static features p4-p7 ARE computed and bound correctly
- Sin_20_y pattern visibility difference between tools under investigation
- Binary weights timestamp (Feb 13 20:36) vs HTML tool (Feb 13 22:12)
- Next: Update HTML tool with canonical binary weights
handoff(Claude): HTML tool weights update pending - base64 encoded
canonical weights ready in /tmp/weights_b64.txt for line 392 replacement.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Diffstat (limited to 'training/target_3')
| -rw-r--r-- | training/target_3/img_000.png | bin | 0 -> 95166 bytes |
| -rw-r--r-- | training/target_3/img_001.png | bin | 0 -> 58701 bytes |
| -rw-r--r-- | training/target_3/img_002.png | bin | 0 -> 73800 bytes |
| -rw-r--r-- | training/target_3/img_003.png | bin | 0 -> 70826 bytes |
| -rw-r--r-- | training/target_3/img_004.png | bin | 0 -> 92114 bytes |
| -rw-r--r-- | training/target_3/img_005.png | bin | 0 -> 62103 bytes |
| -rw-r--r-- | training/target_3/img_006.png | bin | 0 -> 110813 bytes |
| -rw-r--r-- | training/target_3/img_007.png | bin | 0 -> 65858 bytes |
8 files changed, 0 insertions, 0 deletions
diff --git a/training/target_3/img_000.png b/training/target_3/img_000.png Binary files differnew file mode 100644 index 0000000..4af98ef --- /dev/null +++ b/training/target_3/img_000.png diff --git a/training/target_3/img_001.png b/training/target_3/img_001.png Binary files differnew file mode 100644 index 0000000..4b23d2c --- /dev/null +++ b/training/target_3/img_001.png diff --git a/training/target_3/img_002.png b/training/target_3/img_002.png Binary files differnew file mode 100644 index 0000000..7bd6c10 --- /dev/null +++ b/training/target_3/img_002.png diff --git a/training/target_3/img_003.png b/training/target_3/img_003.png Binary files differnew file mode 100644 index 0000000..7d99923 --- /dev/null +++ b/training/target_3/img_003.png diff --git a/training/target_3/img_004.png b/training/target_3/img_004.png Binary files differnew file mode 100644 index 0000000..9a1db32 --- /dev/null +++ b/training/target_3/img_004.png diff --git a/training/target_3/img_005.png b/training/target_3/img_005.png Binary files differnew file mode 100644 index 0000000..cef51f8 --- /dev/null +++ b/training/target_3/img_005.png diff --git a/training/target_3/img_006.png b/training/target_3/img_006.png Binary files differnew file mode 100644 index 0000000..f45d727 --- /dev/null +++ b/training/target_3/img_006.png diff --git a/training/target_3/img_007.png b/training/target_3/img_007.png Binary files differnew file mode 100644 index 0000000..0446ffa --- /dev/null +++ b/training/target_3/img_007.png |
