FAME-Net

This folder contains the implementation of the paper Frequency-Adaptive Pan-Sharpening with Mixture of Experts . Please note that the training datasets are not currently public.

Getting Started

Note: please generate the mask label by dct first, and put them into the dataset folder; refer to the Data class in data/dataset.py

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

In order to run this implementation, you need to have the following software and libraries installed:

• Python 3.7 or later
• PyTorch 1.6 or later
• CUDA (if using GPU)
• NumPy
• Matplotlib
• OpenCV
• PyYAML

Installing

You can install the necessary packages using pip:

pip install torch numpy matplotlib opencv-python pyyaml

Configuration

Before training the model, you need to configure the following options in the option.yaml file:

• log_dir: the directory to store the training log files.
• checkpoint: the directory to store the trained model parameters.
• data_dir_train: the directory of the training data.
• data_dir_eval: the directory of the evaluation data.

Training the Model

To train the model, you can run the following command:

python main.py

Testing the Model

To test the trained WINet model, you can run the following command:

python test.py
python py-tra/demo_deep_methods.py

Configuration

The configuration options are stored in the option.yaml file. Here is an explanation of each of the options:

algorithm

• algorithm: The model for training

Logging

• log_dir: The location where the log files will be stored.

Model Weights

• checkpoint: The location to store the model weights.

Training Data

• data_dir_train: The location of the training data.
• data_dir_eval: The location of the test data.

Pretrain

• pretrained: Whether to use a pretrained model.
• pre_sr: The location of the pretrained model.
• pre_folder: The location where the pretrained models are stored.

Testing

• algorithm: The algorithm to use for testing.
• type: The type of testing, either test or eval.
• data_dir: The location of the test data.
• source_ms: The source of the multi-spectral data.
• source_pan: The source of the panchromatic data.
• model: The location of the model to use for testing.
• save_dir: The location to save the test results.

Data Processing

• upscale: The upscale factor.
• batch_size: The size of each batch.
• patch_size: The size of each patch.
• data_augmentation: Whether to use data augmentation.
• n_colors: The number of color channels.
• rgb_range: The range of the RGB values.
• normalize: Whether to normalize the data.

Training Hyperparameters

• schedule.lr: The learning rate.
• schedule.decay: The learning rate decay.
• schedule.gamma: The learning rate decay factor.
• schedule.optimizer: The optimizer to use, either ADAM, SGD, or RMSprop.
• schedule.momentum: The momentum for the SGD optimizer.