Large-Scale Bidirectional Training for Zero-Shot Image Captioning.
Taehoon KimMark MarsdenPyunghwan AhnSangyun KimSihaeng LeeAlessandra SalaSeung Hwan KimPublished in: CoRR (2022)
Keyphrases
- image classification
- classifier training
- single image
- image analysis
- image data
- image content
- multiscale
- image retrieval
- input image
- image features
- keypoints
- template matching
- low level
- segmentation method
- image representation
- image pixels
- similarity measure
- vector field
- image matching
- spatial information
- test images
- image regions
- image segmentation
- graph cuts
- edge detection
- high resolution
- training data
- real world
- energy function
- training examples
- feature points
- supervised learning
- motion estimation
- pixel values
- million images