Combining image-level and object-level inference for weakly supervised object recognition. Application to fisheries acoustics.
Riwal LefortRonan FabletImen KarouiJean-Marc BoucherPublished in: ICIP (2009)
Keyphrases
- object level
- object class
- pixel level
- weakly supervised
- superpixels
- object recognition
- bounding box
- image features
- object detection
- input image
- multiscale
- low level
- image representation
- partial occlusion
- object detectors
- segmentation method
- segmentation algorithm
- natural images
- single image
- image classification
- image retrieval
- image data
- object classes
- computer vision
- viewpoint
- high resolution
- image content
- image fusion
- region of interest
- image segmentation
- target object
- image patches
- conditional random fields
- keypoints
- object categories
- object segmentation
- appearance model
- semi supervised
- high level
- d objects
- test images
- training data
- higher level