A1519
Title: Optical character recognition enhancement for manufacturing instrument panel numbers with a new object detection loss
Authors: SeokHwan Hong - Inha University (Korea, South) [presenting]
Donghyeon Yu - Inha University (Korea, South)
Abstract: Optical character recognition is widely used in various fields, such as license plate recognition and text extraction from images. In particular, number recognition performs well when the target area is known and fixed. However, unlike license plate recognition, recognizing manufacturing instrument panel numbers remains challenging when the recognition areas vary in location and size. We propose a new loss function for object detection to enhance optical character recognition of manufacturing instrument panel numbers. The proposed loss function enables the object detection method to identify the smallest recognition area that contains all manufacturing instrument panel numbers. Using real images obtained from manufacturing factories, we show that the proposed object detection loss function improves the optical character recognition model for instrument panel numbers.
