System and techniques for automated mesh retopology
Described herein are a system and techniques for performing partially or fully automatic retopology of an object model. In some embodiments, the techniques may involve categorizing and/or segmenting an object model into a number of regions. 3D data in each region may then be compared to 3D data in corresponding regions for a number of similar object models in order to identify a closest matching corresponding region. The techniques may also involve identifying a set of edges stored in relation to each closest matching corresponding region for each region of an object model. Each set of edges may be conformed to the 3D data of its corresponding region. Once conformed, the sets of edges may be compiled into a cage for the object model, from which a mesh may be generated.
1. A method of retopologizing an object model comprising:
maintaining a plurality of retopologized object models, each respective retopologized model of the retopologized object models associated with a respective mesh data, each respective mesh data composed of a plurality of regions which make up the respective retopologized model;
receiving a first object model comprising 3D data associated with an object;
segmenting the first object model into the plurality of regions by assigning, to each of the plurality of regions, separate portions of the 3D data of the first object model;
identifying, for each of the plurality of regions, a closest matching corresponding region from the plurality of regions which make up at least one second object model of the retopologized object models;
determining, for each closest matching corresponding region, mesh data associated with that closest matching corresponding region; and
generating, from the mesh data for each closest matching corresponding region, a retopologized object model by combining the mesh data determined for each closest matching corresponding region.
2. The method of claim 1 , wherein the first object model is a high-resolution object model, and wherein the retopologized object model is a low-resolution object model.
3. The method of claim 1 , wherein the mesh data associated with that closest matching corresponding region comprises a set of edges associated with the closest matching corresponding region from the plurality of regions which make up the at least one second object model.
4. The method of claim 1 , wherein the plurality of retopologized object models are maintained for an object category, the method further comprising identifying the object category as being associated with the object, wherein the plurality of regions are determined based on the object category.
5. The method of claim 4 , wherein the category is identified using one or more machine learning techniques.
6. The method of claim 1 , wherein the plurality of retopologized object models are stored in an object model database.
7. The method of claim 6 , wherein the mesh data associated with the closest matching corresponding region comprises a set of edges associated with the closest matching corresponding region from the plurality of regions which make up the at least one second object model, and wherein the sets of edges are stored in a separate database in relation to the plurality of retopologized object models stored in the object model database.
8. A computing system comprising:
a processor; and
a memory including instructions that, when executed with the processor, cause the computing system to, at least:
maintain a plurality of retopologized object models;
maintain a plurality of mesh data corresponding to the plurality of retopologized object models, each respective mesh data of the plurality of mesh data composed of a plurality of regions which make up a respective retopologized model;
receive a first object model associated with an object;
assign, to each of the plurality of regions, separate portions of the first object model;
identify, for each of the plurality of regions, a closest matching corresponding region from the plurality of regions which make up at least one second object model of the plurality of retopologized object models;
determine, for each closest matching corresponding region, mesh data associated with that closest matching corresponding region; and
generate, from the mesh data for each closest matching corresponding region, a retopologized object model by combining the mesh data determined for each closest matching corresponding region.
9. The computing system of claim 8 , wherein the mesh data associated with the closest matching corresponding region comprises a set of edges associated with the closest matching corresponding region from the plurality of regions which make up the at least one second object model, and wherein the set of edges has been generated by an artist or other user of the system.
10. The computing system of claim 8 , wherein the first object model is segmented into each of the plurality of regions using a machine learning technique.
11. The computing system of claim 8 , wherein the mesh data include a set of edges and wherein the instructions further cause the computing system to conform each of the sets of edges to the first object model.
12. The computing system of claim 11 , wherein conforming each of the sets of edges to the first object model comprises snapping one or more vertices of the sets of edges to one or more vertices of the first object model.
13. The computing system of claim 8 , wherein the mesh data includes edges that correspond to the exterior of the first object model.
14. The computing system of claim 8 , wherein the first object model is a high-resolution mesh generated from point cloud data collected for the object.