Weakly supervised training of universal visual concepts for multi-domain semantic segmentation.
Petra BevandicMarin OrsicIvan GrubisicJosip SaricSinisa SegvicPublished in: CoRR (2022)
Keyphrases
- weakly supervised
- semantic segmentation
- object categories
- object detectors
- object class
- training examples
- superpixels
- topic models
- relation extraction
- object detection
- semi supervised
- object classes
- training set
- information extraction
- small number
- conditional random fields
- image content
- training data
- segmentation algorithm
- object recognition
- image set
- named entities
- object segmentation
- bounding box
- domain specific
- supervised learning
- input image
- active learning