Paper

https://arxiv.org/abs/2409.03179

Code File Descriptions

File	Description
dataset.py	Definition of the Dataset class for training and evaluation, with dataset paths also specified here.
evaluator.py	Code for evaluating the model during training.
inference.py	Code for inferring Super-Resolution (SR) results.
losses.py	Definitions of the loss functions.
test_metrics.py	Code for testing Image Quality Assessment (IQA) metrics.
train_3_loss.py	Implementation of MOBOSR with settings identical to ESRGAN, but employing multi-objective Bayesian optimization to dynamically adjust loss weights during the training process.
train_all_loss.py	Implementation of MOBOSR using all losses.
train_origin.py	Standard implementation of ESRGAN, also utilized during the pre-training phase of MOBOSR.
utils.py	Various utilities.

Pre-trained Model Weights

Model	Download Link
Our-a	Download
Our-b	Download
Our-c	Download
Generator at Pre-train phase	Download
Discriminator at Pre-train phase	Download