diff options
Diffstat (limited to 'doc/CNN_V2_WEB_TOOL.md')
| -rw-r--r-- | doc/CNN_V2_WEB_TOOL.md | 13 |
1 files changed, 10 insertions, 3 deletions
diff --git a/doc/CNN_V2_WEB_TOOL.md b/doc/CNN_V2_WEB_TOOL.md index 8c661b2..25f4ec7 100644 --- a/doc/CNN_V2_WEB_TOOL.md +++ b/doc/CNN_V2_WEB_TOOL.md @@ -10,7 +10,8 @@ Browser-based WebGPU tool for validating CNN v2 inference with layer visualizati **Working:** - ✅ WebGPU initialization and device setup -- ✅ Binary weight file parsing (.bin format) +- ✅ Binary weight file parsing (v1 and v2 formats) +- ✅ Automatic mip-level detection from binary format v2 - ✅ Weight statistics (min/max per layer) - ✅ UI layout with collapsible panels - ✅ Mode switching (Activations/Weights tabs) @@ -24,6 +25,10 @@ Browser-based WebGPU tool for validating CNN v2 inference with layer visualizati - ✅ Mip level selection (p0-p3 features at different resolutions) **Recent Changes (Latest):** +- Binary format v2 support: Reads mip_level from 20-byte header +- Backward compatible: v1 (16-byte header) → mip_level=0 +- Auto-update UI dropdown when loading weights with mip_level +- Display mip_level in metadata panel - Code refactoring: Extracted FULLSCREEN_QUAD_VS shader (reused 3× across pipelines) - Added helper methods: `getDimensions()`, `setVideoControlsEnabled()` - Improved code organization with section headers and comments @@ -70,9 +75,11 @@ Browser-based WebGPU tool for validating CNN v2 inference with layer visualizati ### Key Components **1. Weight Parsing** -- Reads binary format: header (16B) + layer info (20B×N) + f16 weights +- Reads binary format v2: header (20B) + layer info (20B×N) + f16 weights +- Backward compatible with v1: header (16B), mip_level defaults to 0 - Computes min/max per layer via f16 unpacking -- Stores `{ layers[], weights[], fileSize }` +- Stores `{ layers[], weights[], mipLevel, fileSize }` +- Auto-sets UI mip-level dropdown from loaded weights **2. CNN Pipeline** - Static features computation (RGBD + UV + sin + bias → 7D packed) |