A Lightweight Multi-Scale Refinement Network for Gastrointestinal Disease Classification

This is the implementation of the paper "A Lightweight Multi-Scale Refinement Network for Gastrointestinal Disease Classification"

Requirements

-Python 3.11.11

-PyTorch version: 2.6.0

-CUDA version: 12.4

-CUDA device: Tesla T4

-TensorBoard version: 2.18.0

-NumPy version: 1.26.4

-scikit-learn version: 1.2.2

-timm version: 1.0.15

Datasets

Six gastrointestinal datasets were used in this paper as follows:

Kvasir-v1 : https://datasets.simula.no/kvasir/

GastroVision : https://datasets.simula.no/gastrovision/

Kvasir-Capsule : https://datasets.simula.no/kvasir-capsule/

Kvasir-v2 : https://datasets.simula.no/kvasir/

Hyper-Kvasir : https://datasets.simula.no/hyper-kvasir/

WCEBleedGen https://zenodo.org/records/10156571

../                                  
├── data/               
|   ├── data_loader_kvasir_v1.py
│   |   ├── kvasir_v1/
│   |   |   ├── train/
│   |   │   │   ├── NORMAL
|   │   │   │   ├── ...
│   |   |   │   └── Dyed Resection Margins
│   |   |   └── test/
│   |   │   |   ├── NORMAL
|   │   │   |   ├── ...
│   |   |   |   └── Dyed Resection Margins
|   ├── data_loader_kvasir_v2.py
│   |   ├── kvasir_v2/
│   |   |   ├── train/
│   |   │   │   ├── NORMAL
|   │   │   │   ├── ...
│   |   |   │   └── Dyed Resection Margins
│   |   |   └── test/
│   |   │   |   ├── NORMAL
|   │   │   |   ├── ...
│   |   |   |   └── Dyed Resection Margins
|   ├── data_loader_Kvasir_Capsule.py
│   |   ├── Kvasir_Capsule/
│   |   |   ├── train/
│   |   │   |   ├── Ampulla of vater
│   |   │   |   ├──...
│   |   |   |   └── Ulcer
│   |   │   └── test/
│   |   │   |   ├── Ampulla of vater
│   |   │   |   ├──...
│   |   |   |   └── Ulcer
|   ├── data_loader_Hyper_Kvasir.py
│   |   ├── Hyper_Kvasir/
│   |   |   ├── train/
│   |   │   |   ├── barretts
│   |   │   |   ├── ...
│   |   |   |   └── z-line
│   |   |   └── test/
│   |   │   |   ├── barretts
│   |   │   |   ├── ...
│   |   |   |   └── z-line
|   ├── data_loader_Gastrovision.py
│   |   ├── Gastrovision/
│   │   │   ├── train/
│   │   │   │   ├── Accessory tools
│   │   │   │   ├── ...
│   │   │   │   └── Ulcer    
│   │   │   └── test/
│   │   │   │   ├── Accessory tools
│   │   │   │   ├── ...
│   │   │   │   └── Ulcer 
|   ├── data_loader_WCEBleedGen.py
│   |   ├── WCEBleedGen/
│   |   |   ├── train/
│   |   │   |   ├── bleeding
│   |   |   |   └── non-bleeding
│   |   |   └── test/
│   |   │   |   ├── bleeding
│   |   |   |   └── non-bleeding
├── model/
|   ├── backbone_utils.py
|   ├── CBAM.py
|   └── CROSS_ATTENTION.py
|   └── transformer.py
|   └── MSR.py
|   └── FocalLoss.py
├── README.md  
├── dataloader_selection.py
├── Train.py
├── Test.py
├── supplementaries

Training

to train the model use the following command:

python Train.py --bsz 32

          --lr 1e-4
          --niter 60
          --dataset "name_of_the_selected_dataset"
          --num_classes "class_numbers_of_the_selected_dataset"
          --data_dir "your_dataset_directory"
          --logpath "your_experiment_name"

Testing

to test the model use the following command: python Test.py --bsz 32

          --dataset "name_of_the_selected_dataset"
          --num_classes "class_numbers_of_the_selected_dataset"
          --data_dir "your_dataset_directory"
          --weights "your_weights_directory"

Arguments

• --bsz: Batch size for training.

• --lr: Learning rate for the optimizer.

• --niter: Number of training iterations (epochs).

• --data_dir: Directory where the dataset is located.

• --logpath: Directory to save the best model checkpoint.

• --weights: Directory to use the best model checkpoint.

  Citation

If you use this repository in your work, please cite the following paper: