⚠️ Code, comments and documentation are written in Spanish as part of my university coursework.
Asistente Mágico de Escritura is an Android app for children that uses a TensorFlow Lite neural network to recognize handwritten letters in real time, developed as the final project for the Artificial Neural Networks course at UANL.
Young children often struggle with correctly forming letters, and traditional correction methods depend on constant teacher or parent supervision. This app offers an interactive alternative: kids draw uppercase letters on screen, and a trained neural network classifies their handwriting instantly, giving immediate visual feedback.
The model was trained using the A-Z Handwritten Alphabets dataset and deployed as a .tflite file integrated directly into the Android app. The app was designed with a playful space-themed UI to keep young users engaged.
- Touch drawing canvas where children trace uppercase letters (A–Z)
- Real-time letter recognition using a TensorFlow Lite model running on-device
- Softmax probability output across all 26 classes
- Bitmap preprocessing pipeline — normalization, cropping, centering, and 28×28 rescaling
- "Nueva letra" button to randomize the target letter
- Fully offline — no data collected or sent externally
This project is licensed under the MIT License.