Make a Long Image Short: Adaptive Token Length for Vision Transformers.
Qiqi ZhouYichen ZhuPublished in: ECML/PKDD (2) (2023)
Keyphrases
- image data
- multiscale
- single image
- input image
- image classification
- image content
- image features
- visual perception
- image retrieval
- edge detection
- template matching
- image analysis
- computer vision
- region of interest
- lighting conditions
- image set
- image pixels
- low level vision
- test images
- segmentation method
- similarity measure
- image segmentation
- image processing
- vector field
- natural images
- markov random field
- high resolution
- image structure
- pixel values
- image synthesis