Semi‐supervised Mesh Segmentation and Labeling
Boosting
DOI:
10.1111/j.1467-8659.2012.03217.x
Publication Date:
2012-10-02T15:50:30Z
AUTHORS (4)
ABSTRACT
Abstract Recently, approaches have been put forward that focus on the recognition of mesh semantic meanings. These methods usually need prior knowledge learned from training dataset, but when size dataset is small, or meshes are too complex, segmentation performance will be greatly effected. This paper introduces an approach to and labeling which incorporates imparted by both segmented, labeled meshes, unsegmented, unlabeled meshes. A Conditional Random Fields (CRF) based objective function measuring consistency labels faces, neighbouring faces proposed. To implant information we add conditional entropy into function. With entropy, not convex hard optimize, so modify Virtual Evidence Boosting (VEB) solve semi‐supervised problem efficiently. Our yields better results than those only use limited especially many exist. The reduces overall system cost as well human labelling required during training. We also show combining outperforms using either type alone.
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES (23)
CITATIONS (34)
EXTERNAL LINKS
PlumX Metrics
RECOMMENDATIONS
FAIR ASSESSMENT
Coming soon ....
JUPYTER LAB
Coming soon ....