Founder, ΩmegaZNRJ Business Solutions — Ticketeer™ (AI model training for handwritten ticket scanning)
Image
Fine-tuningTrained an AI model (CNN) using raw and synthetic ticket images to recognize and extract handwritten ticket characters with high accuracy. The work involved creating or using datasets for model training and applying optimal training strategies to improve validation performance and generalization. The model was validated to achieve a character accuracy target and then used in a production-like ticket scanning workflow. • Trained CNN with raw and synthetic data for handwritten ticket character recognition. • Applied optimal training strategies to reach strong validation character accuracy. • Ensured generalization performance for real-world handwritten tickets. • Integrated the trained model into a ticket scanning and processing system for downstream data management.