Live In-Browser Webcam Playground & PCA Lab

Experience RCRE's zero-copy CV filter and ML inference pipeline directly in your browser. Take high-resolution photos saved directly to your filesystem, and test real-time PCA image compression with rate-distortion benchmarks.

Synthetic Moving Colorbars FPS: 60.0 (16.6 ms) GBDT: High Contrast (98.4%)

Pipeline Controls

Gaussian Sigma (σ) 1.2
Edge Intensity Threshold 35

Memory & Trace

1. Ingest Tensor [480, 640, 3] 900 KB
2. Grayscale Tensor [480, 640] 300 KB
3. Gaussian Blur Tensor 300 KB
4. Sobel Edges (dx, dy) 600 KB
5. GBDT Decision Inference < 1 μs
Linear Arena Allocation: 3.2 MB / 32 MB (10%)
Zero heap re-allocations in loop.

🔬 PCA Image Compression Test Lab

Test RCRE's zero-dependency Principal Component Analysis (RPCA) spatial block codec. Compress photos into orthogonal low-dimensional subspace bases with 8-bit uniform quantization, inspect learned eigen-basis patches, evaluate rate-distortion curves (PSNR vs. Bitrate), and export bitstreams matching the C23 CLI engine (rcre-pca).

Original Reference
PCA Reconstructed (K=12)
Drag divider left/right or hover over image to inspect pixel-level fidelity
Original Reference Frame
Reconstructed Image (K=12)
Residual Magnitude |I - Î|:
0 (Exact) Low Med High (≥50)
Top principal component eigenvectors W ∈ ℝK × D reshaped into spatial filter patches. Observe how PCA learns DC luminance, directional gradients, and high-frequency textures directly from image pixel statistics.
PSNR Fidelity
-- dB
Ready
Compression Ratio
-- ×
--% space savings
Mean Squared Error
--
L2 residual norm
Bitstream Size
-- KB
-- bpp
Compute Latency
-- ms
Fit + Quant + Recon

Compression Settings

Components Retained (K) 12
Matches RCRE uint8_t min/max .rpca bitstream specification.

Rate-Distortion Benchmark

Sweep K values from 2 to 64 to evaluate PSNR vs. Bitrate curves (identical to rcre-pca --benchmark).

📸
Photo Saved to Filesystem
rcre_photo.png