Doodle Classification

Real-time classification of user-drawn doodles!

This project classifies user-drawn doodles in real-time using a Keras model trained on the Google Doodle dataset. Users can draw on a web interface built with p5.js and HTML, and the Flask framework connects the user interface to the Python backend for classification.

Demo:

Highlights:

• Keras Model: Trained to classify doodles into 5 categories.
• Real-time Classification: Users draw, and the model predicts instantly.
• User Interface (p5.js): Enables interactive drawing experience.
• 92% Accuracy: Achieved through training and optimization.

Setup Instructions:

Install Dependencies:

pip install -r requirements.txt

Modify Labels:

The project uses an array named drawing_lable in the training script train.py to define doodle class labels. Update this array with the specific categories you want your model to classify.

Train the Model:

If you want to train a new model, run the training script:

python train.py

Modal Parameters:

Accuracy vs. Loss Graphs:

Deploy the Model:

• Copy the Model: If you trained a new model, copy the model.pkl file from the project directory to the deployed_app folder.
• Modify Labels (Flask): run app.py file to start the flask app on localhost:5000.