CEREALS - Cost-Effective REgion-based Active Learning for Semantic Segmentation.
Radek MackowiakPhilip LenzOmair GhoriFerran DiegoOliver LangeCarsten RotherPublished in: BMVC (2018)
Keyphrases
- cost effective
- semantic segmentation
- active learning
- street scenes
- conditional random fields
- superpixels
- scene classification
- weakly supervised
- label transfer
- low cost
- object categories
- image segmentation
- training examples
- semi supervised
- learning algorithm
- labeled data
- pascal voc
- object class
- object classes
- multiscale
- cost effectiveness
- supervised learning
- unlabeled data
- learning process
- object recognition
- long range
- training set
- pairwise
- image understanding
- object detection
- data sets
- relevance feedback
- higher order
- image set
- small number
- segmentation algorithm
- natural images
- image classification