Authors: Andrii Denysov

Affiliation: Independent researcher

Email: trolmark@gmail.com

Description: This method formulates the text detection task as one of identifying particular keypoints which are subsequently utilized to delineate the boundaries of curved text regions. A bottom-up approach is employed in this technique, similar to that commonly used for pose-estimation tasks, where all keypoint locations in the image are determined and subsequently grouped together according to geometric rules to form a polygon which defines the curved text area.