VD-SAN: Visual-Densely Semantic Attention Network for Image Caption Generation.
Xinwei HeYang YangBaoguang ShiXiang BaiPublished in: Neurocomputing (2019)
Keyphrases
- low level
- visual features
- visual perception
- input image
- image content
- image data
- visual information
- image retrieval
- single image
- visual appearance
- image features
- visual concepts
- web images
- visually similar
- image analysis
- image segmentation
- image regions
- image collections
- semantic space
- semantic labels
- auto annotation
- image classification
- image representation
- edge detection
- semantic content
- region of interest
- high resolution
- feature points
- multiscale
- visual data
- visual cues
- visual similarity
- low level visual features
- semantic concepts
- low level features
- visual attention
- caption text
- semantic information
- keypoints
- high level
- image processing
- human observers
- visual patterns
- peer to peer
- visual content
- spatial relations
- test images
- segmentation method