VAFMNet

This repository is a lightweight super-resolution training and testing framework built upon BasicSR.

It is designed for convenient development, training, and evaluation of custom image super-resolution models.

---

Project Structure

├── archs/                  # Super-resolution model architectures
├── options/                # YAML configuration files
│   ├── train/
│   └── test/
├── train.py                # Training script
├── test.py                 # Testing script
├── datasets/
├── models/
├── utils/
└── README.md

---

Environment Setup

Clone Repository

git clone https://github.com/MinGTu1/VAFMNet.git
cd VAFMNet

Create Environment

conda create -n basicsr python=3.10
conda activate basicsr

Install Dependencies

pip install torch torchvision
pip install -r requirements.txt

Or install BasicSR manually:

pip install basicsr

---

Dataset Preparation

Organize datasets as follows:

datasets/
├── train/
│   ├── HR/
│   └── LR/
|          X2/
|          X3/
|          X4/
├── val/
│   ├── HR/
│   └── LR/
|          X2/
|          X3/
|          X4/

Dataset paths can be modified in YAML files under options/.

---

Training

Training configuration files are stored in:

options/train/

Run training with:

python train.py -opt options/train/VAFMNet_DIV2K_100w_x2SR.yml

Example:

python train.py -opt options/train/VAFMNet_DIV2K_100w_x2SR.yml

---

Testing

Testing configuration files are stored in:

options/test/

Run testing with:

python test.py -opt options/test/benchmark_VAFMN_x2.yml

Example:

python test.py -opt options/test/benchmark_VAFMN_x2.yml

---

Custom Model

All custom super-resolution architectures should be placed in:

archs/

Example:

archs/
├── VAFMNet_arch.py

After adding a new architecture, register it properly so that it can be called by the YAML configuration.

---

YAML Configuration

All experiment settings are managed through YAML files, including:

• Network structure
• Dataset paths
• Training hyperparameters
• Optimizer settings
• Loss functions
• Validation settings

---

Results

Training logs, checkpoints, and visual results are automatically saved during training/testing.

Typical outputs include:

experiments/
results/

---

Acknowledgements

This project is built upon the excellent framework:

• BasicSR

---

License

This project is released under the MIT License.