SRGAN

This project is an implementation based on the SRGAN project with some modifications. I provides model weights trained for 100 epochs with a 4x upscaling factor on the VOC2012 dataset.

Requirments

pip install requirements.txt

Datsets

Train, Val Dataset

The train and val datasets are sampled from VOC2012. Train dataset has 16700 images and Val dataset has 425 images. Download the datasets from here(Google Drive Link). Then put the training dataset into the data/DIV2K_train_HR folder and place the validation dataset into the data/DIV2K_valid_HR folder.

Usage

Create two empty folders and name them epochs and statistics respectively.
Train

python train.py

optional arguments:
--crop_size                   training images crop size [default value is 88]
--upscale_factor              super resolution upscale factor [default value is 4](choices:[2, 4, 8])
--num_epochs                  train epoch number [default value is 100]

Test Single Image

python test_image.py

optional arguments:
--upscale_factor              super resolution upscale factor [default value is 4]
--test_mode                   using GPU or CPU [default value is 'GPU'](choices:['GPU', 'CPU'])
--image_name                  test low resolution image name
--model_name                  generator model epoch name [default value is netG_epoch_4_100.pth]

Train results

train & val curve

An example of the model trained for 100 epochs on the validation set.

Upscale Factor = 4
The leftmost column is the low-resolution image obtained by interpolation using the BICUBIC method. The middle column is the original high-resolution image, and the rightmost column is the super-resolution image reconstructed using the model.


The outputs of the validation dataset on the model trained for 100 epochs are all located in the training_results\SRF_4 folder.