Speech-Image Semantic Alignment Does Not Depend on Any Prior Classification Tasks.
Masood S. MortazaviPublished in: INTERSPEECH (2020)
Keyphrases
- image features
- image data
- input image
- image content
- image retrieval
- image segmentation
- image classification
- semantic web
- test images
- single image
- image analysis
- high resolution
- image alignment
- edge detection
- region of interest
- image representation
- template matching
- image pixels
- speech recognition
- feature points
- multiscale
- hough transform
- semantic information
- low level
- pixel values
- visual concepts
- spatial information
- image reconstruction
- visual information
- visual features
- prior information
- semantic similarity
- image collections
- automatic speech recognition
- high level