图像超分辨率（super resolution）

方法一：EDSR、WDSR、SRGAN
方法二

方法一：不重新训练的话只能导出4倍

创建环境

$ git clone https://github.com/krasserm/super-resolution.git
$ conda env create -f environment.yml
$ conda activate sisr

下载预训练的权重
weights-edsr-16-x4.tar.gz 地址1 地址2
weights-wdsr-b-32-x4.tar.gz 地址1 地址2
weights-srgan.tar.gz 地址1 地址2

test for srgan

from model.srgan import generator
from utils import load_image, plot_sample
from model import resolve_single
import matplotlib.pyplot as plt
import tensorflow as tf

model = generator()
model.load_weights('weights/srgan/gan_generator.h5')

lr = load_image('demo/xyz.png')
sr = resolve_single(model, lr)

img_numpy=sr.numpy()
# 保存方式1
from PIL import Image
im = Image.fromarray(img_numpy)
im.save("demo/your_file.jpeg")
# 保存方式2
import matplotlib
matplotlib.image.imsave('demo/name.png', img_numpy)