mlx-community/HVI-CIDNet-Generalization-fp32
Image-to-Image • 1.98M • Updated
Apple-MLX HVI-CIDNet exposure correction (CVPR 2025, MIT). Gate on input exposure: these re-expose toward a target luma.
Note Recommended default — least damage when the exposure gate mis-fires (23.37 dB on an already-correct exposure).
Note Headline LOL-v1 model. Strongest lift on severely underexposed input (+14.50 dB at 0.15x exposure).
Note Multi-exposure training. Aggressive auto-exposure normalizer — will DARKEN a bright image by design.