ReLaText: Exploiting visual relationships for arbitrary-shaped scene text detection with graph convolutional networks.
Chixiang MaLei SunZhuoyao ZhongQiang HuoPublished in: Pattern Recognit. (2021)
Keyphrases
- text detection
- arbitrary shaped
- urban scenes
- connected components
- outdoor scenes
- d scene
- object recognition
- text information
- video analysis
- visual data
- visual information
- video sequences
- spatial relations
- low level
- urban environments
- scene understanding
- visual features
- complex scenes
- aerial images
- input image
- spatial relationships
- binary images
- single image
- vanishing points
- image sequences
- arbitrary shape
- complex background
- dynamic scenes
- image based rendering
- multi band
- object detection