Low-Light Image Enhancement with AI (Zero-DCE)

In this video, I demonstrate low-light image enhancement using the Zero-DCE deep learning method (Deep Curve Estimation). I run the pretrained model on three different input types: (1) the authors’ example datasets included with the repository (DICM and LIME), (2) random low-light photos uploaded into Google Colab, and (3) my own raw sensor-like data stored as .npy arrays. For the .npy files, I first convert raw data to an RGB image using a simple ISP pipeline (demosaic, white balance, color correction, and gamma), then apply Zero-DCE to enhance visibility. The result is shown side-by-side (Original | Enhanced) so you can clearly see improvements in brightness and detail. To access the codes and data, use the following link: https://github.com/mortezmaali/Zero-D...