UDC-Baseline

Baseline code for Under-Display-Camera problem

Model

1. UNet (in UNet.py)

   classic deep learning structure

2. DE_UNet (in DE_UNet.py)

   image restoration model in CVPR_2021_paper: Image Restoration for Under-Display Camera

Some Results

TOLED results with UNet trained 400 epochs (Centercrop patches, 256*256) and DE_UNet trained 400 epochs(same input)

Problems to be solved

1. 由于有些输入与label之间的亮度有差距，有些没有，结果呈现出许多亮度高于原图的问题。
2. 可以见到的是网络普遍可以解决blur和noise问题，但是复原问题仍旧不尽人意。