Boosting OCR Accuracy with Diverse Textual Image Collection

Client's Challenge & Our Solution

A major tech company sought to enhance its Optical Character Recognition (OCR) and text recognition technology and needed a diverse dataset of textual images collected exclusively from iOS devices. The company was very focused on the diversity in the dataset. We set together and identified all possible diversity scenarios and drafted a detailed data collection plan.

FutureBeeAI stepped in to gather high-quality images, including both printed and handwritten text, from various iOS devices such as iPhones, iPads, and iPods. We ensured that the dataset included diverse text formats like invoices, flyers, letters, forms, business cards, menus, storefronts, etc. The dataset was further diverse in terms of lighting conditions, capture angle, and background.

Outcome & Features:

Gathered over 100,000 textual images, including handwritten notes and printed documents, from various iOS devices.

Ensured diversity with images captured in different lighting, angles, and document types to simulate real-world scenarios.

Delivered a fully annotated dataset that significantly enhanced the client's OCR accuracy across iOS platforms.