Update README.md

README.md

@@ -3,81 +3,48 @@ title: Brain Tumor Detection
 emoji: 🐢
 colorFrom: yellow
 colorTo: green
-sdk: gradio
-sdk_version: 5.42.0
+sdk: streamlit
 app_file: app.py
 pinned: false
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
-🧠 Brain Tumor Detection (Multi-Modal)
-This is a Gradio-based web app for detecting brain tumors using a multi-modal deep learning model. The app supports two modes:
+# 🧠 Brain Tumor Detection (CNN Only)
 
-Image-Only Mode: Uses a CNN model to predict based solely on MRI images.
-Multi-Modal Mode: Combines MRI images (processed via CNN) and tabular patient data (processed via MLP) for enhanced predictions.
+This is a **Streamlit-based web app** for detecting brain tumors using a **CNN model** trained on MRI images. The app only uses **image input** for prediction — tabular data is not required.
 
-Features
+## Features
 
-Upload an MRI image (JPG format).
-Optionally input patient data (Mean, Variance, Standard Deviation, Entropy, Skewness, Kurtosis, Contrast, Energy, ASM, Homogeneity, Dissimilarity, Correlation, Coarseness, PSNR, SSIM, MSE, DC).
-Automatically selects the appropriate model:
-CNN-only model if only an image is provided.
-Multi-modal model if both image and tabular data are provided.
+- Upload a brain MRI image (`.jpg`, `.jpeg`, `.png`).
+- Predicts **Benign** or **Malignant** tumors.
+- Uses a CNN trained on grayscale MRI images resized to 128x128 pixels.
+- Lightweight and fast — suitable for deployment on Hugging Face Spaces.
 
+## How to Use
 
-Outputs prediction ("Tumor" or "No Tumor") with confidence score and indicates the model used.
+1. **Upload MRI Image:** Select a brain MRI image (JPG, JPEG, or PNG format).  
+2. **View Prediction:** The app will display the uploaded image and predict:
+   - **Benign**
+   - **Malignant**  
 
-How to Use
+The prediction is displayed directly in the app.
 
-Upload MRI Image: Select a brain MRI image (JPG format).
-Enter Patient Data (Optional): Input numerical values for the 17 features extracted from the image or patient records. Leave blank for image-only prediction.
-Submit: Click the submit button to get the prediction.
+## Model Details
 
-Model Details
+- **CNN Architecture:**  
+  - 3 convolutional layers (32, 64, 128 filters) with ReLU activation  
+  - Max-pooling and dropout layers for regularization  
+  - Final dense layer with sigmoid activation
 
-Multi-Modal Model:
-Architecture:
-CNN branch: 3 convolutional layers (32, 64, 128 filters) with ReLU, max-pooling, and dropout.
-MLP branch: 2 dense layers (64, 32 units) for tabular data.
-Combined features fed to a final dense layer with sigmoid activation.
+- **Input:** Grayscale MRI image, resized to 128x128, normalized to `[0,1]`  
+- **Training Dataset:** Brain Tumor images only  
+- **Metrics:** Binary classification (Tumor vs. No Tumor)
 
+## Installation (For Local Development)
 
-Inputs: MRI image (128x128, grayscale) + 17 tabular features.
+1. Clone the repository.  
+2. Install dependencies:
 
+```bash
+pip install -r requirements.txt
 
-CNN-Only Model:
-Architecture: 3 convolutional layers (32, 64, 128 filters) with ReLU, max-pooling, dropout, and a final dense layer with sigmoid activation.
-Inputs: MRI image (128x128, grayscale).
-
-
-Training Data: Brain Tumor dataset (images) + tabular features from bt_dataset_t3.csv.
-Preprocessing:
-Images: Grayscale, resized to 128x128, normalized to [0,1].
-Tabular: Standardized using StandardScaler.
-
-
-Metrics: Binary classification (Tumor vs. No Tumor).
-
-Installation (For Local Development)
-
-Clone this repository.
-Install dependencies: pip install -r requirements.txt
-Run the app: python app.py
-
-Deployment
-
-Deploy on Hugging Face Spaces.
-Ensure multi_modal_brain_tumor_model.h5, cnn_only_brain_tumor_model.h5, and scaler.pkl are included in the repository.
-
-Notes
-
-Ensure the MRI image is in JPG format and matches the training data format (grayscale, 128x128).
-Tabular inputs are optional. If provided, they must be numeric and correspond to the 17 features used during training.
-The app automatically switches to the CNN-only model if tabular data is not provided.
-Model files (multi_modal_brain_tumor_model.h5, cnn_only_brain_tumor_model.h5) and scaler (scaler.pkl) must be present in the same directory.
-
-Acknowledgments
-
-Built with Gradio and TensorFlow.
-Dataset inspired from Kaggle Brain Tumor features and images.