Progressive Evolution from Single-Point to Polygon for Scene Text

• DOI: 10.48550/arxiv.2312.13778
• Publication Date: 2023-01-01
• Authors: Linger Deng; Mingxin Huang; Xudong Xie; Yuliang Liu; Lianwen Jin; Xiang Bai
• Keywords: Polygon (computer graphics); Spotting; Ground truth; Point in polygon; Single point

The advancement of text shape representations towards compactness has enhanced detection and spotting performance, but at a high annotation cost. Current models use single-point annotations to reduce costs, yet they lack sufficient localization information for downstream applications. To overcome this limitation, we introduce Point2Polygon, which can efficiently transform single-points into compact polygons. Our method uses coarse-to-fine process, starting with creating selecting anchor points based on recognition confidence, then vertically horizontally refining the polygon using optimize its shape. We demonstrate accuracy generated polygons through extensive experiments: 1) By from ground truth points, achieved an 82.0% ICDAR 2015; 2) In training detectors by our method, attained 86% relative (GT); 3) Additionally, proposed Point2Polygon be seamlessly integrated empower spotters generate This integration led impressive 82.5% It is worth mentioning that relies solely synthetic information, eliminating need any manual beyond single points.