Self-training via Metric Learning for Source-Free Domain Adaptation of Semantic Segmentation.
Ibrahim Batuhan AkkayaUgur HaliciPublished in: CoRR (2022)
Keyphrases
- metric learning
- domain adaptation
- semantic segmentation
- semi supervised
- semi supervised learning
- weakly supervised
- co training
- labeled data
- target domain
- unlabeled data
- multiple sources
- conditional random fields
- superpixels
- pairwise
- transfer learning
- supervised learning
- distance metric
- active learning
- generative model
- object categories
- unsupervised learning
- multi view
- text categorization
- object classes
- distance function
- multi task
- data points
- dimensionality reduction
- data sets
- image classification
- feature space
- prior knowledge
- support vector machine
- object recognition
- support vector
- feature selection
- machine learning