Zero-shot semantic segmentation via spatial and multi-scale aware visual class embedding.
Sungguk ChaYooseung WangPublished in: Pattern Recognit. Lett. (2022)
Keyphrases
- semantic segmentation
- multiscale
- superpixels
- object categories
- object classes
- street scenes
- spatial relations
- label transfer
- conditional random fields
- scale space
- object detection
- scene classification
- visual features
- weakly supervised
- image classification
- spatial information
- image representation
- natural images
- graphical models
- object class
- object classification
- pixel level
- image processing
- d objects