Neural network pipeline making digital images accessible for color vision deficient users. End-to-end personalized color compensation for CVD. FM100-based 3D psychophysical profiling (θ, C, S) conditions CVDAdaN (ConvNeXt encoder + PLCF decoder), Y-preserving forward in YCbCr. Outperforms CUD-Net & Farup on perceptual alignment. ~22 FPS inference.
computer-vision deep-learning image-processing pytorch color-blindness color-vision-deficiency daltonism cvd-compensation machado-simulation farup-algorithm-application dataset-daltonism farnsworth-munsell fm100-hue vingry-king-smith vingrys
-
Updated
Mar 31, 2026 - Python