ACORT: A compact object relation transformer for parameter efficient image captioning.
Jia Huei TanYing Hua TanChee Seng ChanJoon Huang ChuahPublished in: Neurocomputing (2022)
Keyphrases
- image features
- image regions
- image retrieval
- single image
- multiple objects
- input image
- target object
- image data
- image content
- multiscale
- image classification
- keypoints
- pixel level
- bounding box
- feature points
- image representation
- normalized correlation
- object features
- location and orientation
- image segmentation
- spatial relations
- image analysis
- high resolution
- visual appearance
- object localization
- pixel values
- edge detection
- image matching
- image segments
- three dimensional objects
- foreground and background
- intensity images
- ground plane
- segmentation algorithm
- test images
- lighting conditions
- segmentation method
- spatial relationships
- partial occlusion
- fuzzy logic
- background clutter
- image collections
- d objects
- object shape
- binary codes
- similar objects
- individual objects
- low level
- fault diagnosis
- object tracking