Think as People: Context-Driven Multi-Image News Captioning with Adaptive Dual Attention.
Qiang YangXiaodong WuXiuying ChenXin GaoXiangliang ZhangPublished in: ICASSP (2024)
Keyphrases
- input image
- multiscale
- image data
- image classification
- image features
- image pixels
- image collections
- high resolution
- image analysis
- image content
- similarity measure
- template matching
- single image
- test images
- segmentation method
- hough transform
- pixel values
- contextual information
- news articles
- lighting conditions
- region of interest
- vector field
- computer vision
- image representation
- natural images
- image quality
- multiresolution
- image segmentation