search

tensorlayer / SRGAN

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
https://github.com/tensorlayer/tensorlayerx
3.24k stars 813 forks source link

Feeding Lr_images to the generator separately, without downscaling the HR_Images #223

Open mustaeenqazi opened 3 years ago

mustaeenqazi commented 3 years ago

Hi, Thanks for great work, I want to apply SRGAN, to a images of a model, whose resolution is affected by resizing from high resolution to low resolution as done in train.py LINE 62 " lr_patch = tf.image.resize(hr_patch, size=[96, 96])" I have my own low resolution image folder which i want to feed to the tf.data.Dataset.from_generator, so that the Generator network learns those instead of downsampled Hr_images In the image attached is some changes i made which somehow couldnt work out. it seems the " tf.data.Dataset.from_generator" how_to_feed_lr_separately or "train_ds.map" causes an issue of dtype. Please have a look, any possible solutions would help alot. thanks in advance.