Font conversion for steel product number recognition: A conditioned diffusion model approach

Taehan Lee, Hyeyeon Choi, Bum Jun Kim, Hyeonah Jang, Donggeon Lee, Sang Woo Kim

Abstract: Highlights•Handwritten numbers vary in font, leading to data inconsistency.•Inconsistency causes low accuracy of recognition models.•Font Changer can convert various fonts into a representative font.•Font Changer successfully converts even messy or damaged images.•By ensuring font consistency, recognition accuracy improves.