We support every pixel architecture
Glass AI models can be trained for any sensor: any pixel size from large pixel to sub-micron, any CFA type, and any CFA pattern. Our physics-based approach reconstructs real detail regardless of the underlying pixel technology, achieving clarity that conventional demosaicing pipelines can't match. Integration requires only RAW or binned RAW data from the target sensor.
New sensors ≠ New ISP
Traditional ISPs require extensive hand-tuning for each new sensor configuration. Glass AI does not. The same training framework adapts to new CFA arrangements with no performance penalty, whether you're working with standard, quad, or hex Bayer sensors, or emerging architectures like RGBW and RGB-IR. Future sensor innovation becomes a drop-in integration, not a new pipeline project.
.png)
