Few-Shot Defect Segmentation Leveraging Abundant Normal Training Samples Through Normal Background Regularization and Crop-and-Paste Operation.
Dongyun LinYanpeng CaoWenbing ZhuYiqun LiPublished in: CoRR (2020)
Keyphrases
- training samples
- feature space
- learning algorithm
- training data
- training set
- segmentation algorithm
- high dimensional
- number of training samples
- test sample
- decision function
- decision boundary
- segmentation method
- image segmentation
- supervised learning
- video sequences
- hyperplane
- face images
- high resolution
- high quality
- shape prior
- sample set
- image processing
- computer vision
- data mining