Optical Braille Recognition Using Object Detection Neural Network

Ilya G. Ovodov; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2021, pp. 1741-1748

Optical Braille recognition methods generally rely heavily on a Braille text's geometric structure. They run into problems if this structure is distorted. Thus, they find it difficult to cope with images of book pages taken with a smartphone. We propose an optical Braille recognition method that uses an object detection convolutional neural network to detect whole Braille characters at once. The proposed algorithm is robust to deformations and perspective distortions of a Braille page displayed on an image. The algorithm is suitable for recognizing braille texts captured with a smartphone camera in domestic conditions. It can handle curved pages and images with perspective distortion. The proposed algorithm shows high performance and accuracy compared to existing methods. Additionally, we produced a new dataset containing 240 photos of Braille texts with annotation for each Braille letter. Both the proposed algorithm and the dataset are available at GitHub.