AMLSII_22-23_SN22081179

ELEC 0135 Applied Machine Learning Systems II aimed at solving the NTIRE 2017 super resolution challenge

1. Prerequisites

To Begin with, it is required to clone this project or download to your computer or server. The structure of the folders are as follows:

AMLSII_22-23_SN22081179

• A/
  • datasets.py
• B/
  • datasets.py
• Datasets/
  • DIV2K_train_HR/
  • DIV2K_train_LR_bicubic_X2/
  • DIV2K_train_LR_unknown/
  • DIV2K_valid_HR/
  • DIV2K_valid_LR_bicubic_X2/
  • DIV2K_valid_LR_unknown/
• src/
  • models.py
  • utils.py
• checkpoint_srgan_A.pth.tar
• checkpoint_srgan_B.pth.tar
• checkpoint_srresnet_A.pth.tar
• checkpoint_srresnet_B.pth.tar
• eval_srgan_A.py
• eval_srgan_B.py
• eval_srresnet_A.py
• eval_srresnet_B.py
• super_resolve_A.py
• train_srgan_A.py
• train_srgan_B.py
• train_srresnet_A.py
• train_srresnet_B.py
• super_resolve_A.py

The folder Datasets is not included, you can either download the required Datesets according to the file tree or you can download the zip file here. The checkpoints can also be downloaded here in case you accidentally updated the original one during the training process.

The environment

My advice is to create a new conda environment from the environment.yml file in this repo environment.yml You can simply do it by:

conda env create -f environment.yml

2. How to check the result of this project

If your server or computer is GPU ready

(For the project I used turin.ee.ucl.ac.uk, I checked whether it woeks on this server, I strongly suggest that you test the project on the server) You can simply check by input:

torch.cuda.is_available()

If the output is True, then congratulation that you can start from the training by exectuting the corresponding file to solve the Task A and B. However, you should always train the SRResnet before training the SRGAN as the weights of SRResnet are needed when training SRGAN.

This means for task A the training sequence should be:

train_srresnet_A.py -> train_srgan_A.py

For task B the training sequence should be:

train_srresnet_B.py -> train_srgan_B.py

You can also just check the result by executing the eval funciton

For task A it would be executing:

• eval_srresnet_A.py
• eval_srgan_A.py

For task B it would be executing:

• eval_srresnet_B.py
• eval_srgan_B.py

To generate the image with subplot of upsamapled image and the original image for comparision run:

• super_resolve_A.py