OroLight: Oral Cancer Detection using Deep Learning

OroLight is a web-based deep learning application that classifies oral histopathology images as cancerous or non-cancerous. It integrates a trained InceptionResNetV2 model within a Flask app and provides Grad-CAM visualizations for interoperability.

---

Features

• Upload and classify oral histopathology images
• Uses InceptionResNetV2 for high accuracy
• Includes Grad-CAM to visualize model attention
• Fully styled UI with animations
• Training notebook with graphs and metrics

---

Model Training and Performance

The model is trained on the Oral Cancer Dataset by ZaidPy.

Achieved 90% accuracy using transfer learning and fine-tuning.

---

🔧 Setup Instructions

1. Clone the Repository

git clone https://github.com/Jasmine5220/orolight.git
cd orolight

2. Create a Virtual Environment

python -m venv venv
source venv/bin/activate      # MacOS/Linux
venv\Scripts\activate         # Windows

3. Download the Dataset

1. Go to the dataset page: 🔗 https://www.kaggle.com/datasets/zaidpy/oral-cancer-dataset
2. Download the zip folder of the dataset.

4. Train the Model

Run the training notebook.

This will:

• Load and preprocess images
• Train an InceptionResNetV2-based classifier
• Save the best model as best_model.h5 in orolight folder

5. Move the Model to the Correct Directory

After training completes:

mv best_model.h5 oral_cancer_app/model/

6. Go to Flask App and Install Required Packages

cd oral_cancer_app
pip install -r requirements.txt

Or manually:

pip install flask tensorflow keras opencv-python matplotlib numpy pandas

7. Run the Flask App

python app.py

Visit http://127.0.0.1:5000 in your browser.

---

Example Test Images

Inside Tests/, you’ll find:

• real-test-1.jpg
• real-test-2.jpg

These are sample images(taken in Jabalpur Medical College), you can upload via the web app to test the model.

---

Accuracy Visualization

Training and validation accuracy plotted over epochs.

---

This project was part of EDP (DFP-24).