Exploring Object-Centric and Scene-Centric CNN Features and Their Complementarity for Human Rights Violations Recognition in Images.
Grigorios KalliatakisShoaib EhsanAles LeonardisMaria FasliKlaus D. McDonald-MaierPublished in: IEEE Access (2019)
Keyphrases
- object models
- visual appearance
- images depicting
- static images
- target object
- image features
- objects in cluttered scenes
- image regions
- multiple objects
- geometric information
- object model
- object recognition
- input image
- individual objects
- occluded objects
- three dimensional objects
- intensity images
- complex scenes
- acquired images
- man made objects
- recognizing objects
- real world scenes
- spatial relationships
- feature correspondences
- three dimensional
- viewing angle
- multiple images
- object representation
- background clutter
- object features
- varying illumination conditions
- image set
- textural information
- real objects
- d objects
- test images
- visual scene
- object parts
- keypoints
- wire frame
- feature extraction
- geometric constraints
- point features
- image segments
- object detectors
- scene images
- partial occlusion
- region features
- convolutional neural network
- real world objects
- lighting conditions
- spatial relations
- cad model
- image matching
- object segmentation
- range data
- surface properties
- image sequences
- viewing direction
- spatial information
- visual features
- salient objects
- spatial location
- action recognition
- face detection
- single image
- image classification
- range finder
- object detection
- moving objects
- video sequences
- text localization and recognition