This is modified code of Keras Zero-DCE implementation
https://keras.io/examples/vision/zero_dce/
Zero-Reference Deep Curve Estimation (Zero-DCE) is a method for improving the visibility of images taken in low light conditions using a deep neural network. It does this by estimating an image-specific tonal curve, which is a mathematical function that maps the brightness levels of the original image to new, enhanced brightness levels.This is done without any reference image, hence the term "zero-reference."