[Model Enhancement]: Brain Tumor Detection with Web Application

[TechieSamosa]

Brain Tumor Detection Improvement Task

Description

The existing brain tumor detection project has been implemented on a small dataset with suboptimal accuracy and recall rates. The project is also missing an app.py file.

Proposed Task

• Improve the accuracy and recall rates of the brain tumor detection model.
• Add an app.py file to create a web application for tumor detection.

Please Assign To

@TheNaiveSamosa

Notes

As a beginner, the assigned contributor will attempt to enhance the existing brain tumor detection project.

[abhisheks008]

Looks good to me! Can you please mention the details accourding to the given chart,

✅ To be Mentioned while taking the issue :

• Full name :

• GitHub Profile Link :

• Email ID :

• Approach for this Project :

• What is your participant role? (Mention the Open Source program)

[TechieSamosa]

---

Full name: Aditya Khamitkar

GitHub Profile Link: TheNaiveSamosa

Email ID: thenaivesamosa@gmail.com

Participant Role: Contributor for GirlScript Summer of Code

Approach

1. Loading the Dataset

• Read the training and testing datasets into dataframes.
• Generated file paths with labels for images.

2. Data Preprocessing

• Checked if the training data is balanced.
• Split the testing data into test and validation sets.

3. Image Data Generation

• Created an ImageDataGenerator for training, validation, and testing dataframes.
• Used flow_from_dataframe to convert dataframes to numpy arrays.

4. Model Structure

• Defined the model using Sequential API.
• Added Conv2D, MaxPooling2D, Flatten, and Dense layers.
• Compiled the model using Adamax optimizer and categorical_crossentropy loss.

5. Model Training

• Trained the model using the fit method on the training generator.
• Specified the number of epochs and used the validation generator for validation data.

6. Model Performance

• Plotted training and validation loss and accuracy over epochs.
• Identified the best epoch based on validation loss and accuracy.

7. Model Evaluation

• Evaluated the model on training, validation, and testing data.
• Printed the loss and accuracy for each dataset.

8. Prediction and Evaluation

• Generated predictions using the model on the testing data.
• Calculated the confusion matrix and classification report.
• Saved the model to an HDF5 file.

This approach follows a standard workflow for building, training, and evaluating a deep learning model for image classification. It involves loading and preprocessing data, defining the model architecture, training the model, evaluating its performance, and saving it for future use.

---

[TechieSamosa]

Approach for the Flask application I was thinking of adding. @abhisheks008

---

1. Loading Model and Labels

• Loaded the pre-trained TensorFlow model for brain tumor classification.
• Defined the class labels for different tumor types ('Glioma', 'Meningioma', 'No tumor', 'Pituitary').

2. Preprocessing Image

• Defined a function to preprocess the uploaded image.
• Resized the image to (150, 150) pixels.
• Converted the image to a NumPy array and expanded its dimensions.

3. Converting Image to Base64

• Defined a function to convert the processed image to a base64 string.
• Used io.BytesIO to buffer the image and base64.b64encode to encode it.

4. Flask App Setup

• Initialized a Flask application.
• Loaded the model and class labels.

5. Routes

• Defined a route ('/') to render the index.html template for uploading an image.
• Defined a route ('/predict') to handle the image prediction.
• Extracted the uploaded image, preprocessed it, and made predictions using the model.
• Converted the predicted class to a label and rendered the result.html template with the prediction and the base64 encoded image.

6. Error Handling

• Implemented error handling to catch any exceptions that might occur during image processing or prediction.
• Rendered an error.html template with the error message.

7. Running the Application

• Used the if __name__ == '__main__': block to run the Flask application in debug mode.

This approach sets up a Flask application for predicting brain tumor types based on uploaded images. It handles image preprocessing, model prediction, and error handling, providing a user-friendly interface for interacting with the model.

---

Please do assign this to me.

[abhisheks008]

One more addition, can you make a demonstration video of the application and upload it in the README file. @TheNaiveSamosa

[TechieSamosa]

I'm planning to submit a PR with some changes, but I'd appreciate it if you could review the changes first. The required HTML files are quite basic, and as I'm not very familiar with frontend development, it might take me some time to create the demo video. Could I submit the code first and then work on the demo video once the changes are approved?

Regarding the demo video in the README, I understand it's a requirement, but I'd prefer if you could assign it to me officially before I create the video. I'd like to ensure the changes are approved before investing time in creating the demo.

[abhisheks008]

It'll be better if you push the codes after completing the whole work, as a result it'll be considered as level 3, which will value your time and efforts.

Issue assigned to you @TheNaiveSamosa

[TechieSamosa]

Done. Please review the PR.

[abhisheks008]

The project folder structure should be like this,

Project Dummy
|- Dataset
  |- dataset.csv
  |- README.md
|- Images
  |- img1.png
  |- img2.png
  |- img3.png
|- Model
  |- project_dummy.ipynb
  |- README.md
|- Web App
  |- app.py
  |- templates
  |- demo.mp4
  |- README.md
|- requirements.txt

Hope this will give you a clear perspective of the requirement. @TheNaiveSamosa