On Regularized Losses for Weakly-supervised CNN Segmentation.
Meng TangFederico PerazziAbdelaziz DjelouahIsmail Ben AyedChristopher SchroersYuri BoykovPublished in: CoRR (2018)
Keyphrases
- weakly supervised
- semantic segmentation
- fully supervised
- superpixels
- segmentation algorithm
- topic models
- object class
- relation extraction
- medical images
- semi supervised
- segmentation method
- object segmentation
- level set
- machine learning
- multiscale
- bounding box
- shape prior
- named entities
- information extraction
- domain specific
- viewpoint
- ground truth
- image segmentation
- object categories
- input image
- natural images
- conditional random fields
- image classification
- markov random field
- object detection
- image features
- computer vision