Distinct Class-Specific Saliency Maps for Weakly Supervised Semantic Segmentation.
Wataru ShimodaKeiji YanaiPublished in: ECCV (4) (2016)
Keyphrases
- weakly supervised
- saliency map
- superpixels
- input image
- object segmentation
- object class
- natural images
- visual attention
- object recognition
- multiscale
- human visual system
- object detection
- viewpoint
- object classes
- higher order
- visual features
- image patches
- bounding box
- video frames
- low level features
- object categories
- learning algorithm
- segmentation method
- segmentation algorithm
- appearance model
- partial occlusion
- image classification
- denoising
- image representation
- graph structure
- higher level
- object model
- d objects
- face recognition