2020/05/25 by Mayank Singh, Mayank Kumar Singh, Singh, Mayank Kumar +4 · 1 citation
Computer Science · Engineering · Mathematics · #Algorithm #Artificial intelligence #Artificial neural network #Bounding overwatch #Character (mathematics) #Computation and Language (cs.CL) #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Engineering #Enhanced Data Rates for GSM Evolution #FOS: Computer and information sciences #Graph #Handwritten Text Recognition Techniques #Image (mathematics) #Link (geometry) #Machine Learning (cs.LG) #Mathematics #Minimum bounding box #Multimodal Machine Learning Applications #Natural Language Processing Techniques #Node (physics) #Pattern recognition (psychology) #State (computer science) #Theoretical computer science #Word (group theory) #cs.CL #cs.CV #cs.LG
paper · pdf · doi:10.48550/arxiv.2005.12147
published in arXiv (Cornell University) (Cornell University) · 9 pages
arxiv created 2020/05/25 · openalex publication_date 2020/05/25 · arxiv updated 2020/05/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05
Text detection in scenes based on deep neural networks have shown promising results. Instead of using word bounding box regression, recent state-of-the-art methods have started focusing on character bounding box and pixel-level prediction. This necessitates the need to link adjacent characters, which we propose in this paper using a novel Graph Neural Network (GNN) architecture that allows us to learn both node and edge features as opposed to only the node features under the typical GNN. The main advantage of using GNN for link prediction lies in its ability to connect characters which are spatially separated and have an arbitrary orientation. We show our concept on the well known SynthText dataset, achieving top results as compared to state-of-the-art methods.