Weighting Pseudo-Labels via High-Activation Feature Index Similarity and Object Detection for Semi-Supervised Segmentation.
Prantik HowladerHieu LeDimitris SamarasPublished in: CoRR (2024)
Keyphrases
- object detection
- semi supervised
- image parsing
- partially labeled
- segmented images
- object segmentation
- similarity index
- similarity measure
- image labeling
- fully supervised
- structural similarity
- label propagation
- semi supervised learning
- pairwise
- segmentation algorithm
- object hypotheses
- background subtraction
- medical images
- pairwise constraints
- segmentation method
- unlabeled data
- feature vectors
- level set
- semantic segmentation
- unlabeled samples
- training set
- image segmentation
- object categories
- training data
- multiscale
- feature set
- face detection
- shape prior
- distance measure
- similarity queries
- semi supervised clustering
- multi view
- label information
- training examples
- labeled data
- multiple features
- superpixels
- metric learning
- pedestrian detection
- feature vectors are extracted
- label field
- computer vision
- index structure
- object recognition
- feature space
- labeling problems
- active learning
- segmented regions
- co training
- labeled examples