From 7a05f4d33b611ba1e9b6c68e0d0bd67d6ea011ee Mon Sep 17 00:00:00 2001 From: skal Date: Tue, 10 Feb 2026 21:11:05 +0100 Subject: refactor: Optimize CNN grayscale computation Compute gray once per fragment using dot() instead of per-layer. Pass gray as f32 parameter to conv functions instead of vec4 original. Co-Authored-By: Claude Sonnet 4.5 --- doc/CNN_EFFECT.md | 23 ++++++++++++++--------- doc/CNN_RGBD_GRAYSCALE_SUMMARY.md | 10 ++++++---- 2 files changed, 20 insertions(+), 13 deletions(-) (limited to 'doc') diff --git a/doc/CNN_EFFECT.md b/doc/CNN_EFFECT.md index 4659fd3..22cf985 100644 --- a/doc/CNN_EFFECT.md +++ b/doc/CNN_EFFECT.md @@ -38,7 +38,7 @@ fn cnn_conv3x3_7to4( samp: sampler, uv: vec2, resolution: vec2, - original: vec4, # Original RGBD [-1,1] + gray: f32, # Grayscale [-1,1] weights: array, 36> # 9 pos × 4 out × (7 weights + bias) ) -> vec4 @@ -48,7 +48,7 @@ fn cnn_conv3x3_7to1( samp: sampler, uv: vec2, resolution: vec2, - original: vec4, + gray: f32, weights: array, 9> # 9 pos × (7 weights + bias) ) -> f32 ``` @@ -56,7 +56,7 @@ fn cnn_conv3x3_7to1( **Input normalization:** - **fs_main** normalizes textures once: `(tex - 0.5) * 2` → [-1,1] - **Conv functions** normalize UV coords: `(uv - 0.5) * 2` → [-1,1] -- **Grayscale** computed from normalized RGBD: `0.2126*R + 0.7152*G + 0.0722*B` +- **Grayscale** computed once in fs_main using dot product: `dot(original.rgb, vec3(0.2126, 0.7152, 0.0722))` - **Inter-layer data** stays in [-1,1] (no denormalization) - **Final output** denormalized for display: `(result + 1.0) * 0.5` → [0,1] @@ -250,20 +250,25 @@ Expands to: ```wgsl @fragment fn fs_main(@builtin(position) p: vec4) -> @location(0) vec4 { let uv = p.xy / uniforms.resolution; - let input = textureSample(txt, smplr, uv); // Layer N-1 output - let original = textureSample(original_input, smplr, uv); // Layer 0 input - + let original_raw = textureSample(original_input, smplr, uv); + let original = (original_raw - 0.5) * 2.0; // Normalize to [-1,1] + let gray = dot(original.rgb, vec3(0.2126, 0.7152, 0.0722)); var result = vec4(0.0); if (params.layer_index == 0) { - result = cnn_conv3x3_with_coord(txt, smplr, uv, uniforms.resolution, - rgba_weights_layer0, coord_weights_layer0, bias_layer0); + result = cnn_conv3x3_7to4_src(txt, smplr, uv, uniforms.resolution, + weights_layer0); + result = cnn_tanh(result); + } + else if (params.layer_index == 1) { + result = cnn_conv5x5_7to4(txt, smplr, uv, uniforms.resolution, + gray, weights_layer1); result = cnn_tanh(result); } // ... other layers // Blend with ORIGINAL input (not previous layer) - return mix(original, result, params.blend_amount); + return mix(original_raw, result, params.blend_amount); } ``` diff --git a/doc/CNN_RGBD_GRAYSCALE_SUMMARY.md b/doc/CNN_RGBD_GRAYSCALE_SUMMARY.md index 4c13693..3439f2c 100644 --- a/doc/CNN_RGBD_GRAYSCALE_SUMMARY.md +++ b/doc/CNN_RGBD_GRAYSCALE_SUMMARY.md @@ -20,7 +20,7 @@ Implemented CNN architecture upgrade: RGBD input → grayscale output with 7-cha - **RGBD:** `(rgbd - 0.5) * 2` - **UV coords:** `(uv - 0.5) * 2` -- **Grayscale:** `(0.2126*R + 0.7152*G + 0.0722*B - 0.5) * 2` +- **Grayscale:** `dot(original.rgb, vec3(0.2126, 0.7152, 0.0722))` (computed once, passed as parameter) **Rationale:** Zero-centered inputs for tanh activation, better gradient flow. @@ -48,13 +48,14 @@ Implemented CNN architecture upgrade: RGBD input → grayscale output with 7-cha **Shaders (`/Users/skal/demo/workspaces/main/shaders/cnn/cnn_conv3x3.wgsl`):** 1. Added `cnn_conv3x3_7to4()`: - - 7-channel input: [RGBD, uv_x, uv_y, gray] + - 7-channel input: [RGBD, uv_x, uv_y, gray] (gray passed as parameter) - 4-channel output: RGBD - Weights: `array, 36>` 2. Added `cnn_conv3x3_7to1()`: - - 7-channel input: [RGBD, uv_x, uv_y, gray] + - 7-channel input: [RGBD, uv_x, uv_y, gray] (gray passed as parameter) - 1-channel output: grayscale - Weights: `array, 9>` +3. Optimized: gray computed once in caller using `dot()`, not per-function **Documentation (`/Users/skal/demo/doc/CNN_EFFECT.md`):** 1. Updated architecture section with RGBD→grayscale pipeline @@ -71,7 +72,8 @@ CNNLayerParams and bind groups remain unchanged. 2. Each layer: - Samples previous layer output (RGBD in [0,1]) - Normalizes RGBD to [-1,1] - - Computes UV coords and grayscale, normalizes to [-1,1] + - Computes gray once using `dot()` (fs_main level) + - Normalizes UV coords to [-1,1] (inside conv functions) - Concatenates 7-channel input - Applies convolution with layer-specific weights - Outputs RGBD (inner) or grayscale (final) in [-1,1] -- cgit v1.2.3