On Regularized Losses for Weakly-supervised CNN Segmentation.
Meng TangFederico PerazziAbdelaziz DjelouahIsmail Ben AyedChristopher SchroersYuri BoykovPublished in: ECCV (16) (2018)
Keyphrases
- weakly supervised
- semantic segmentation
- superpixels
- fully supervised
- segmentation algorithm
- object class
- object segmentation
- segmentation method
- image segmentation
- multiscale
- relation extraction
- medical images
- topic models
- semi supervised
- level set
- bounding box
- object recognition
- energy function
- graph cuts
- conditional random fields
- long range
- higher order
- object detectors
- probabilistic model
- training set