Optimized_TumorClassifierResNetSD

A Novel Optimized ResNet-SD Model for Brain Tumor Classification Using Stochastic Depth and Metaheuristic Optimization.

This repository contains code for a deep learning model that detects brain tumors in MRI images. The model is implemented using the ResNet-50 architecture, optimized with stochastic depth and metaheuristics, and trained on a dataset of 5712 images, including the Glioma, Meningioma, Pituitary, and Normal classes.

This is the official Optimized TumorClassifierResNetSD implementation repository with PyTorch.

Optimizied TumorClassifierResNetSD Architecture

Setup

Clone the repository.

!git clone https://github.com/nafiuny/Optimized_TumorClassifierResNetSD.git

%cd Optimized_TumorClassifierResNetSD

!pip install -r requirements.txt

Option1: Download and Preprocess the dataset

If you want to download the dataset and perform the preprocessing yourself, you can use the following code.
The first command will download the dataset from Google Drive, and the second one will process the data into the required format.

!python download_dataset.py

!python preprocess.py \
        --train_dir Brain_Tumor_MRI_Dataset/Training \
        --test_dir Brain_Tumor_MRI_Dataset/Testing \
        --output_dir outputs/data

Option2: Use preprocessed files

If you prefer to use the preprocessed files without running the preprocessing step, you can download the preprocessed files (train_data.pt, val_data.pt, and test_data.pt) using the following code.
Once the preprocessed files are downloaded, you can move directly to training the model.

!python download_preprocessed_data.py

Dataset

Finally, the structure of the 'data' folder where the preprocessed files are placed is as follows.

/outputs/data
├── test_label.pt
├── test_data.pt
├── train_label.pt
├── train_data.pt
├── test_label.pt
├── test_data.pt

Train

Train Optimizied_TumorClassifierResNetSD.
You can modify the values of the input parameters as per your requirements. Alternatively, you can run the script with the default settings without changing any parameters.

!python train.py --model_name "resnet_sd" \
        --checkpoint_name "model_resnet_sd" \
        --train_data_path "outputs/data/train_data.pt" \
        --train_labels_path "outputs/data/train_labels.pt" \
        --val_data_path "outputs/data/val_data.pt" \
        --val_labels_path "outputs/data/val_labels.pt" \
        --num_epochs 200 \
        --lr 0.01 \
        --stochastic_depth1 0.6 \
        --stochastic_depth2 0.7 

Test

In the following command, enter the name of the checkpoint specified in the train step and run it.

!python test.py --model_name "resnet_sd" \
        --checkpoint_path "outputs/models/model_resnet_sd.pth" \
        --test_data_path "outputs/data/test_data.pt" \
        --test_labels_path "outputs/data/test_labels.pt" 

Best Model

To use the best trained model, run the following command.

!python download_best_model_resnet_sd.py

Plot

Run the following code to display the confusion matrix.

from IPython.display import Image, display
print("Displaying plots:")
display(Image("outputs/plots/confusion_matrix_resnet_sd.png"))

Run the following code to show the Train and Validation Loss and Accuracty history .

!python plot_history.py --checkpoint_path="outputs/models/model_resnet_sd.pth"

from IPython.display import Image, display
print("Displaying plots:")
display(Image("outputs/plots/Train_loss_model_resnet_sd.png"))
display(Image("outputs/plots/Train_acc_model_resnet_sd.png"))

Test Model quickly

To quickly detect a tumor from an MRI image, run the following command.

!python application_test.py --image_path /content/drive/MyDrive/n4.png --checkpoint_path outputs/models/best_model_resnet_sd.pth

checkpoint_path is the address of the most recently checkpointed
image_path is the address of the MRI file you want to detect