fadhilelrizanda/swin_base_alzheimer_mri

Model Card for Alzheimer MRI Image Classification using Swin Base Transformer

Developed by: Fadhil Elrizanda (github:fadhilelrizanda)

Model Details

The model is finetuned from https://github.com/microsoft/Swin-Transformer, which using configuration microsoft/swin-base-patch4-window7-224. The model is already finetuned for the Falah/Alzheimer_MRI dataset

Direct Use

-- Exploration model especially for swin transformer in medical domain

-- Visualization of swin transformer using gradcam

Performance and Visualization

Grad Cam

Confusion Matrix

Classification Report

Class	Precision	Recall	F1-Score	Support
Mild_Demented	0.981	0.895	0.936	172
Moderate_Demented	1.000	0.800	0.889	15
Non_Demented	0.958	0.983	0.970	634
Very_Mild_Demented	0.959	0.963	0.961	459
Accuracy			0.962	1280
Macro Avg	0.975	0.910	0.939	1280
Weighted Avg	0.962	0.962	0.961	1280

How to use

git clone https://huggingface.co/fadhilelrizanda/swin_base_alzheimer_mri
cd swin_base_alzheimer_mri

import torch
from model import SwinCustomClassification  

# Initialize the model
model = SwinCustomClassification()
model.load_state_dict(torch.load("swin_base_best.pt"))
model.eval()

Full code for exploration and reproduction : https://github.com/fadhilelrizanda/image_classification_alzheimer_mri