Three-dimensional segmentation of digital models utilizing soft classification geometric tuning

Patent metadata

Field	Value
Publication number	US-10706554-B2
Application number	US-201715487813-A
Country	US
Kind code	B2
Filing date	Apr 14, 2017
Priority date	Apr 14, 2017
Publication date	Jul 7, 2020
Grant date	Jul 7, 2020

Abstract

The present disclosure includes methods and systems for identifying and manipulating a segment of a three-dimensional digital model based on soft classification of the three-dimensional digital model. In particular, one or more embodiments of the disclosed systems and methods identify a soft classification of a digital model and utilize the soft classification to tune segmentation algorithms. For example, the disclosed systems and methods can utilize a soft classification to select a segmentation algorithm from a plurality of segmentation algorithms, to combine segmentation parameters from a plurality of segmentation algorithms, and/or to identify input parameters for a segmentation algorithm. The disclosed systems and methods can utilize the tuned segmentation algorithms to accurately and efficiently identify a segment of a three-dimensional digital model.

First claim

We claim:

1. A computer-implemented method of selecting and manipulating segments of three-dimensional digital models, comprising:
determining a soft classification of a three-dimensional digital model, from a plurality of soft classifications, by analyzing features of the three-dimensional digital model based on a plurality of training digital models and training soft classification categories;
identifying a plurality of segmentation algorithms for segmenting digital models;
utilizing the soft classification of the three-dimensional digital model determined from the plurality of soft classifications to select a segmentation algorithm to apply to the three-dimensional digital model from the plurality of segmentation algorithms;
receiving an indication of a selection of a portion of the three-dimensional digital model; and
identifying a segment of the three-dimensional digital model corresponding to the selection utilizing the segmentation algorithm selected utilizing the soft classification corresponding to the three-dimensional digital model.

2. The method of claim 1 , wherein determining the soft classification of the three-dimensional digital model comprises utilizing a soft classification algorithm; and
further comprising training the soft classification algorithm prior to receiving the indication of the selection by:
providing the training digital models to the soft classification algorithm;
for each training digital model, utilizing the soft classification algorithm to predict at least one soft classification category corresponding to the training digital model; and
for each training digital model, comparing the at least one predicted soft classification category with a training soft classification category corresponding to the training digital model.

3. The method of claim 1 , wherein determining the soft classification comprises determining, for each soft classification category of a plurality of soft classification categories, a probability that that the three-dimensional digital model corresponds to the soft classification category.

4. The method of claim 3 , wherein determining the segmentation algorithm comprises determining a correspondence between the soft classification category of the plurality of soft classification categories and the segmentation algorithm from the plurality of segmentation algorithms.

5. The method of claim 4 , wherein determining the segmentation algorithm comprises comparing a first probability, from the soft classification, that the three-dimensional digital model corresponds to the soft classification category with a second probability, from the soft classification, that the three-dimensional digital model corresponds to a second soft classification category.

6. The method of claim 3 , wherein determining the segmentation algorithm from the plurality of segmentation algorithms further comprises:
selecting a first segmentation algorithm from the plurality of segmentation algorithms based on a first probability from the soft classification; and
selecting a second segmentation algorithm different from the first segmentation algorithm from the plurality of segmentation algorithms based on a second probability from the soft classification.

7. The method of claim 6 , further comprising:
generating a first segmentation parameter for the three-dimensional digital model utilizing the segmentation algorithm;
generating a second segmentation parameter for the three-dimensional digital model utilizing the second segmentation algorithm; and
generating a mixed segmentation parameter based on the first probability, the second probability, the first segmentation parameter, and the second segmentation parameter.

8. The method of claim 7 , wherein identifying the segment of the three-dimensional digital model corresponding to the selection utilizing the segmentation algorithm comprises identifying the segment of the three-dimensional digital model utilizing the mixed segmentation parameter.

9. The method of claim 3 , wherein determining the segmentation algorithm comprises:
comparing a probability threshold with a first probability, from the soft classification, that the three-dimensional digital model corresponds to a first classification category; and
based on a determination that the first probability exceeds the probability threshold, selecting the segmentation algorithm.

10. The method of claim 1 , wherein:
determining the segmentation algorithm comprises determining an input parameter based on the soft classification; and
utilizing the segmentation algorithm comprises utilizing the input parameter determined based on the soft classification to identify the segment of the three-dimensional digital model.

11. A system for selecting segments of three-dimensional digital models, comprising:
one or more memories storing a set of instructions comprising:
a soft classification algorithm trained to generate soft classifications of three-dimensional digital models from a plurality of soft classifications, the soft classifications comprising probabilities that a given three-dimensional digital model corresponds to one or more soft classification categories in a set of soft classification categories;
a plurality of segmentation algorithms, wherein each segmentation algorithm corresponds to a soft classification category from the set of soft classification categories; and
a three-dimensional digital model comprising a plurality of vertices; and
at least one computing device storing instructions thereon, that, when executed by the at least one computing device, cause the system to:
determine a soft classification of the three-dimensional digital model, from the plurality of soft classifications, utilizing the soft classification algorithm;
utilize the soft classification of the three-dimensional digital model determined from the plurality of soft classifications to select a segmentation algorithm to apply to the three-dimensional digital model from the plurality of segmentation algorithms;
receive an indication of a selection of a portion of the three-dimensional digital model; and
identify a segment of the three-dimensional digital model corresponding to the selected portion of the three-dimensional digital model utilizing the segmentation algorithm selected based on the soft classification of the three-dimensional digital model.

12. The system of claim 11 , further comprising instructions that, when executed by the at least one computing device, cause the system to: determine the soft classification by determining, for each soft classification category in the set of soft classification categories, a probability that that the three-dimensional digital model corresponds to the soft classification category.

13. The system of claim 12 , further comprising instructions that, when executed by the at least one computing device, cause the system to:
select the segmentation algorithm from the plurality of segmentation algorithms based on a first probability from the soft classification;
select a second segmentation algorithm from the plurality of segmentation algorithms based on a second probability from the soft classification;
generate a first segmentation parameter for the three-dimensional digital model utilizing the segmentation algorithm;
generate a second segmentation parameter for the three-dimensional digital model utilizing the second segmentation algorithm;
generate a mixed segmentation parameter based on the first segmentation parameter, the second segmentation parameter, and mixture coefficients; and
apply the segmentation algorithm to the three-dimensional digital model by identifying the segment

Assignees

• Adobe Inc

Inventors

• Kim Vladimir
• Hertzmann Aaron
• Yumer Mehmet

Classifications

• G06V10/26

  Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion · CPC title

• G06V30/248

  involving plural approaches, e.g. verification by template match; Resolving confusion among similar patterns, e.g. "O" versus "Q" (G06V30/242 takes precedence) · CPC title

• G06V20/64

  Three-dimensional [3D] objects · CPC title

• G06T2200/04

  involving 3D image data · CPC title

• G06T2207/20081

  Training; Learning · CPC title