ReLaText: Exploiting Visual Relationships for Arbitrary-Shaped Scene Text Detection with Graph Convolutional Networks.
Chixiang MaLei SunZhuoyao ZhongQiang HuoPublished in: CoRR (2020)
Keyphrases
- text detection
- arbitrary shaped
- urban scenes
- connected components
- outdoor scenes
- visual data
- scene understanding
- d scene
- spatial relations
- text information
- visual information
- video analysis
- video sequences
- structure from motion
- image sequences
- spatial relationships
- point cloud
- complex scenes
- single image
- visual features
- web images
- image based rendering
- feature points
- input image