Two-Stage Training for Improved Classification of Poorly Localized Object Images.
Sravanthi BondugulaGang QianAllison BeachPublished in: ECCV Workshops (5) (2020)
Keyphrases
- image classification
- image data
- image regions
- rigid body
- lighting conditions
- deformable objects
- training set
- multiple objects
- image analysis
- supervised learning
- bounding box
- image database
- input image
- support vector
- image set
- image features
- objects represented
- background clutter
- region of interest
- complex scenes
- image matching
- visual appearance
- training samples
- hyperspectral images
- segmentation algorithm
- spatially localized
- three dimensional
- discriminative classifiers
- image retrieval
- individual objects
- classification accuracy
- text classification
- segmentation method
- surface shape
- image annotation
- spatial relationships
- test images
- object models
- normalized correlation
- training phase
- object detectors
- real objects
- feature selection
- target object
- object tracking
- edge detection
- d objects
- feature vectors
- feature space
- object recognition
- decision trees
- training process
- object segmentation
- keypoints
- computer vision
- pigmented skin lesions