The main goal of the project is to create a machine learning model that can accurately scan handwritten text from images or documents and convert it into digital text. This tool has wide applications in digitizing notes, automating documentation, improving accessibility, and reducing human errors caused by illegible handwriting. The system will be trained to recognize individual characters in multiple writing styles (e.g., cursive, italic, monospaced), progressing to recognizing complete words and finally full sentences. The project serves as a practical implementation of OCR (Optical Character Recognition) powered by CNNs and Recurrent Neural Networks (RNNs), allowing students to explore both image processing and sequential data learning.
The project is organized into a twelve-week schedule, starting with foundational knowledge and moving into model development and testing. In the initial weeks, students will study the underlying concepts of CNNs and their application to image recognition tasks. A dataset of handwritten text in different formats will be collected or created to train the model. Necessary machine learning libraries such as TensorFlow or PyTorch will be installed and configured. Students will begin building a basic model framework and progressively train it to recognize individual letters, then groups of characters, and ultimately full words and sentences.
Subsequent phases will involve improving the model's performance with additional datasets and applying techniques to increase accuracy. The team will also integrate a camera module or use scanned images for testing real-world handwritten inputs. The project will conclude with documentation, extensive testing for accuracy and error handling, and a final presentation. Ethical development, originality, and consistent coding standards are expected throughout the project duration.
