A user-friendly web app that recognizes handwritten digits using a Convolutional Neural Network (CNN) trained on the MNIST dataset. Built with TensorFlow, Flask, and JavaScript, this app allows users to draw digits on a canvas and get instant predictions from the ML model. The front-end features a pastel-themed interface for a more delightful experience.
- Draw digits on a responsive canvas (mouse & touch supported)
- Clear and Predict buttons for ease of interaction
- Real-time prediction using a trained CNN model
- Light and aesthetic pastel-colored UI
- Python (Flask for backend)
- TensorFlow / Keras (for training and inference)
- HTML / CSS / JavaScript (frontend)
- PIL / NumPy (image processing)
A pastel-themed canvas with clear controls and minimal distractions for the users.
This project is open source and available under the MIT License.