Triangle-Reward Reinforcement Learning: A Visual-Linguistic Semantic Alignment for Image Captioning.
Weizhi NieJiesi LiNing XuAn-An LiuXuanya LiYongdong ZhangPublished in: ACM Multimedia (2021)
Keyphrases
- reinforcement learning
- low level
- input image
- multiscale
- image classification
- image data
- image features
- visual perception
- natural language
- auto annotation
- image retrieval
- image collections
- single image
- image content
- image alignment
- image representation
- high resolution
- high level
- visual information
- visual features
- visual concepts
- segmentation method
- semantic labels
- image analysis
- image segmentation
- function approximation
- visually similar
- visual appearance
- web images
- visual input
- machine learning
- edge detection
- semantic space
- visual attributes
- image processing
- low level visual features
- visual cues
- semantic concepts
- visual data
- spatial relations
- low level features