Optical Flow Training under Limited Label Budget via Active Learning.
Shuai YuanXian SunHannah Halin KimShuzhi YuCarlo TomasiPublished in: CoRR (2022)
Keyphrases
- active learning
- optical flow
- labeling effort
- active learner
- training set
- training examples
- supervised learning
- label noise
- selective sampling
- annotation effort
- active learning framework
- sample selection
- random sampling
- motion field
- image sequences
- learning strategies
- least squares
- flow field
- machine learning
- online learning
- labeled data
- unlabeled data
- motion estimation
- unlabeled instances
- labeled instances
- batch mode
- limited memory
- training process
- test set
- training samples
- feature tracking
- optical flow estimation
- optical flow computation
- training phase
- generalization error
- semi supervised learning
- qualitative and quantitative evaluation
- crowd sourced
- motion analysis
- dense optical flow
- learning process
- learning algorithm