SF-FSDA: Source-Free Few-Shot Domain Adaptive Object Detection with Efficient Labeled Data Factory.
Han SunRui GongKonrad SchindlerLuc Van GoolPublished in: CoLLAs (2023)
Keyphrases
- labeled data
- target domain
- domain adaptation
- transfer learning
- unlabeled data
- semi supervised
- semi supervised learning
- object detection
- active learning
- training data
- source domain
- supervised learning
- cross domain
- text classification
- multiple sources
- data points
- semi supervised classification
- labeled examples
- domain specific
- prior knowledge
- labeled training data
- class labels
- learning tasks
- co training
- real world
- label propagation
- learning algorithm
- computer vision
- object recognition
- pairwise
- similarity measure
- text categorization
- training samples
- collaborative filtering