JuZiSYJ / ncnn-picture-enhancement

a dehaze / picture enhancement demo using ncnn

ncnn-picture-enhancement

A simple demo to run dehaze / underwater model in Android

Curretly Implemented:

• AOD-Net：dehaze
• UWGAN :underwater enhancement
• GLADNet :low light enhancement

Need to do

• git clone
• download ncnn libs for Android
• put the libs in src/main/cpp/ncnn-android-vulkan
• models are already provided in assets
• apk(https://pan.baidu.com/s/1c6ArUPwIeWiHinANNqfuLg) vr23

Results

• AON-Net

• UNet(model is already update, old version is pic/Unet.jpg)

• GLADNet

Time cost

using Xiaomi, 845cpu

	cpu(*4)	gpu
AODNet(800*500)	200ms
UNet(400*550)		150ms
GLADNet(640*480)		600ms

Reference

• ncnn
• AOD-Net
• UWGAN