Unsupervised confidence-based refinement (IMAGE)
Caption
An example of this phenomenon is shown in the figure, where nongeometric local features prove inadequate for capturing the spatial proximity of the contacting appendages. By filtering out these types of features during the training and testing phases, better performance and more accurate alignment results can be achieved. The proposed SSGHM overcomes the enduring issue of feature ambiguity in dense-shape correspondence by harnessing the power of the Gromov–Hausdorff distance metric for 3D alignments. SSGHM identifies and selects the most informative features that lead to a more robust and accurate correspondence between shapes, enabling real-time confidence-based refinement and yielding state-of-the-art results that significantly enhance convergence times.
Credit
Beijing Zhongke Journal Publising Co. Ltd.
Usage Restrictions
Credit must be given to the creator.
License
CC BY