Bottom-up and Top-down Object Inference Networks for Image Captioning.
Yingwei PanYehao LiTing YaoTao MeiPublished in: ACM Trans. Multim. Comput. Commun. Appl. (2023)
Keyphrases
- image regions
- single image
- multiscale
- bounding box
- region of interest
- image data
- input image
- image features
- image content
- keypoints
- normalized correlation
- target object
- test images
- image analysis
- multiple objects
- image representation
- complex scenes
- random fields
- spatial relationships
- spatial relations
- high resolution
- partial occlusion
- lighting conditions
- individual objects
- three dimensional objects
- background clutter
- pixel level
- position and orientation
- segmentation method
- spatial information
- object tracking
- d objects
- particle filter
- image retrieval
- pixel values
- belief networks
- intensity images
- foreground and background
- object localization
- low level
- surface shape
- image set
- object features
- image segmentation