Modulating and attending the source image during encoding improves Multimodal Translation.
Jean-Benoit DelbrouckStéphane DupontPublished in: CoRR (2017)
Keyphrases
- input image
- image data
- image analysis
- image content
- image segmentation
- multiscale
- template matching
- image features
- image retrieval
- single image
- image classification
- segmentation algorithm
- image regions
- hough transform
- image matching
- multimodal image registration
- decoding process
- image pixels
- image collections
- high resolution
- test images
- vector field
- segmentation method
- edge detection
- image representation
- feature points
- low level
- keypoints
- region of interest
- binary images
- mutual information