Reducing Annotation Effort by Identifying and Labeling Contextually Diverse Classes for Semantic Segmentation Under Domain Shift.
Sharat AgarwalSaket AnandChetan AroraPublished in: CoRR (2022)
Keyphrases
- semantic segmentation
- street scenes
- annotation effort
- object classes
- active learning
- conditional random fields
- superpixels
- scene classification
- label transfer
- object categories
- weakly supervised
- domain specific
- object class
- d objects
- object segmentation
- unlabeled data
- training set
- object recognition
- multiscale
- image understanding
- biologically inspired
- image set
- data mining
- text mining
- small number
- decision trees
- e learning
- computer vision