Back to browse
Foveon – Bayer to Foveon X3, learned, Mac App using deep learning

Foveon – Bayer to Foveon X3, learned, Mac App using deep learning

by coolwulf·Jun 28, 2026·2 points·1 comment

AI Analysis

●●●BangerBig BrainNiche GemSlick

U-Net with 3-layer pixel-stack encoding emulates Foveon X3 sensors better than Lightroom presets.

Strengths
  • Modified U-Net injects 1D pixel-stack encoding at bottleneck — architecture matches the physical sensor problem
  • Trained on matched Bayer → Merrill scene pairs, not synthetic data or style transfer hacks
  • Ships as signed 33 MB macOS app with UI sliders, not just a research paper or Colab notebook
Weaknesses
  • Apple Silicon only, no Windows or Linux support limits the photographer audience significantly
  • Niche appeal — only matters if you already care about the Foveon Merrill look
Category
Target Audience

Photographers, computational photography enthusiasts, Mac users with Apple Silicon

Similar To

Topaz Photo AI · DxO PureRAW · Adobe Lightroom AI Enhance

Similar Projects