AI-Powered Handwritten Notes Summarization

Handwritten notes are valuable in learning and working, but reading long documents involves substantial manual labor.This paper presents a Python research prototype that combines handwriting recognition, human correction, and automatic summarization.A Streamlit interface accepts images and multipage PDFs, applies optional preprocessing, segments pages into text lines, and extracts editable text using a pretrained TrOCR model.Users can correct the transcription before generating a length-controlled summary with a local Qwen2.5 instruction-tuned language model.Long inputs are processed through token-aware chunking and hierarchical consolidation.The application includes TXT export and an evaluation harness for recognition error, summary overlap, processing time, and human review.A proposed extension addresses English, Telugu, Hindi, Tamil, Kannada, and Malayalam through script-appropriate recognition and translation.The supplied implementation supports the English workflow, while multilingual operation and empirical performance remain to be validated.The contribution is a transparent integration and evaluation framework that exposes recognition errors before they propagate into summaries.