Abstract: A camera system and method capture image data with a camera, a data storage device electrically connected to the camera and configured to store the video data and/or a communication device electrically connected to the camera and configured to communicate the image data to a system receiver located remote from the camera. The system receiver may be located onboard a vehicle such that an operator can carry the camera off board the vehicle and communicate the image data back to the vehicle, when performing, for example, work on the vehicle or inspecting the vehicle or the environs of the vehicle.

Abstract: A computer-based method for hybrid ray tracing to generate photorealistic reflections of objects on a non-planar reflective surface. Each triangle of the reflective surface is handled separately for secondary rays to generate its own physically correct reflection. A final gathering of partial reflections results in an aggregated non-planar surface reflection.

Abstract: Systems, apparatuses and methods may provide a way to subdivide a patch generated in graphics processing pipeline into sub-patches, and generate sub-patch tessellations for the sub-patches. More particularly, systems, apparatuses and methods may provide a way to diverge tessellation sizes to a configurable size within an interior region of a patch or sub-patches based on a position of each of the tessellations. The systems, apparatuses and methods may determine a number of tessellation factors to use based on one or more of a level of granularity of one or more domains of a scene to be digitally rendered, available computing capacity, or power consumption to compute the number of tessellation factors.

Abstract: Systems and methods for image based location estimation are described. In one example embodiment, a first positioning system is used to generate a first position estimate. Point cloud data describing an environment is then accessed. A two-dimensional surface of an image of an environment is captured, and a portion of the image is matched to a portion of key points in the point cloud data. An augmented reality object is then aligned within one or more images of the environment based on the match of the point cloud with the image. In some embodiments, building façade data may additionally be used to determine a device location and place the augmented reality object within an image.

Abstract: Aspects of the technology described herein provide a control interface for manipulating a 3-D graphical object within a virtual drawing space. The control can be activated by selecting a graphical object or objects. When multiple objects are selected, the manipulations can occur as a group. In one aspect, the manipulations occur around the centroid of the 3-D graphical object, or groups of objects. The manipulations can include rotation, size adjustment, and positional adjustment within the virtual drawing space. The control interface comprises a visible mesh that helps the user position the object in space relative to other objects.

Abstract: A computing system may include a mesh access engine and a mesh repair engine. The mesh access engine may be configured to access a boundary representation of an object design, the boundary representation including a defect. The mesh repair engine may be configured to repair the boundary representation, including by converting the boundary representation into a spatial decomposition representation of the object design and converting the spatial decomposition representation of the object design back into a boundary representation form to obtain a repaired boundary representation of the object design.

Abstract: An audio segmentation method based on an attention mechanism is provided. The audio segmentation method according to an embodiment obtains a mapping relationship between an “inputted text” and an “audio spectrum feature vector for generating an audio signal”, the audio spectrum feature vector being automatically synthesized by using the inputted text, and segments an inputted audio signal by using the mapping relationship. Accordingly, high quality can be guaranteed and the effort, time, and cost can be noticeably reduced through audio segmentation utilizing the attention mechanism.

Abstract: A ray-tracing system configured to perform intersection testing, comprising: a tester module for testing rays for intersection with a volume, the tester module being configured to receive a packet of one or more rays to be tested for intersection with the volume, wherein the tester module comprises: a first set of one or more testers configured to perform intersection testing at a first level of precision to provide intersection testing results, wherein for a first type of the intersection testing result from the first set of one or more testers intersection testing does not need to be reperformed at a second level of precision greater than the first level of precision, and for a second type of the intersection testing result from the first set of one or more testers intersection testing is to be reperformed at the second level of precision; and a second set of one or more testers configured to perform intersection testing at the second level of precision; wherein the tester module is configured to: allocate

Abstract: In one embodiment, a writing data generation method is for generating writing data used by a multi-charged particle beam writing apparatus. The writing data generation method includes referring to library data in which a vertex sequence including a plurality of vertices is registered, and extracting a portion of an outer line of a figure contained in design data, the portion corresponding to the vertex sequence, and representing the extracted portion by information which identifies the vertex sequence and information which indicates a connection method for the plurality of vertices of the vertex sequence, and generating the writing data.

Abstract: A system, apparatus and method for generating or modifying a display of objects using a “drag and drop” metaphor and coupling that to the processing of data records in a database. This results in a modification of the underlying records as needed to reflect the changes to the arrangement or ordering of content representing a set of objects.

Abstract: An operation object includes a polyhedron. The polyhedron includes characteristic surfaces formed by cutting each vertex of a cube at different angles. An information processing apparatus acquires a normal vector of each surface from a polarized image thereof. Further, the information processing apparatus identifies the characteristic surface by defining, as a characteristic amount, an angle formed by an average vector Na of normal vectors N1, N2, and N3 of reference surfaces that orthogonally intersect each other and a normal vector N0 of the characteristic surface surrounded by the reference surfaces. The specification result and the normal vectors thereof are used to acquire the position and orientation of the operation object.

Abstract: The disclosed embodiments relate to a representation of a structure/space as a collection of interrelated partitions whereby manipulation of the partitions causes an automated propagation of the manipulation through the other interrelated partitions as a function of the relationships therebetween. Each partition is defined by one or more vertices, the location(s) of which is/are specified either relative to an origin or relative to, i.e. as a function of, or dependent upon, the location of another vertex within the partition (intra-partition) or within another partition (inter-partition). Where the location of a vertex is moved, the change is propagated implicitly to all other vertices whose location is related, i.e. specified as a function of the modified vertex.

Abstract: A computer-automated method is presented for segmenting image data for an organ of a subject, where the organ is a tubular structure. The method includes: receiving image data representing a volume of the subject, such that the image data includes the organ; generating a centerline through the organ; determining location of an outer wall of the tube within the image data, where the location of the inner wall is determined using the centerline; determining location of an outer wall of the tube within the image data, where the location of the outer wall is determined using the inner wall; and computing a measure of the organ from the image data.

Abstract: Systems and methods for reducing hops associated with a head mounted display are described. The head mounted display includes a communications circuit for receiving and transmitting interactive media associated with a game program via a network. The interactive media is processed by the game cloud system and streamed directly to the communications circuit of the head mounted display. The head mounted display further includes a user input circuit for receiving an action from a user to generate an input, which includes position and motion detected by the user input circuit. The head mounted display includes a game processing circuit for decoding the interactive media received from the network. The game processing circuit drives a portion of interactivity associated with the game program. The portion of interactivity is generated based on the input.

Abstract: A method of controlling the order in which primitives generated during tessellation are output by the tessellation unit involves sub-dividing a patch, selecting one of the two sub-patches which are formed by the sub-division and tessellating that sub-patch until no further sub-division is possible before tessellating the other (non-selected) sub-patch. The method is recursively applied at each level of sub-division. Patches are output as primitives at the point in the method where they do not require any further sub-division. The selection of a sub-patch is made based on the values of one or more flags and any suitable tessellation method may be used to determine whether to sub-divide a patch. Methods of controlling the order in which vertices are output by the tessellation unit are also described and these may be used in combination with, or independently of, the method of controlling the primitive order.

Abstract: A unit that divides an image into a mesh shape and acquires depth information and an X-coordinate and a Y-coordinate for vertices of each grid, a unit that transforms vertex data into three-dimensional space coordinates expressed by a predetermined coordinate system, a unit that calculates line segment vectors and a normal vector that constitute each grid based on the three-dimensional space coordinates of each vertex, a vector management part that assigns indexes for distinguishing unique vectors from each other to the respective unique vectors, and generates a vector list in which the unique vectors and the indexes are respectively associated with each other, and a unit that stores therein indexes of the vectors that constitute each grid, while being associated with the grid.

Abstract: A method for real-time volumetric 3-D reconstruction of an object uses at least one depth sensor camera. In a preparation step, voxels are collected in a reconstructed scene depending on a new depth map frame, and the collected voxels are cached in order to perform an update of the reconstructed scene. In an integration step, the collected and cached voxels of the preparation step are updated with a newly captured depth map frame. However, the preparation step, and the integration step are separated from each other so that both steps can be carried out in parallel, and at the same time.

Abstract: A computer-implemented method for generating roof reports. In one embodiment, a user begins by entering an address into an application and clicking a “request” button. A 3D model of the structure at the address is prepared. The application may then proceed with manual or automatic roof extraction. For manual roof extraction, an analyst may outline roof facets to create a roof model. For automatic roof extraction, the application may use machine learning or another technique to automatically extract and then simplify roof facets. In either case, the application may then present the roof model to a reviewer for inspection. The reviewer may then accept or reject the roof model. If the reviewer rejects the roof model, an analyst may correct the roof model and resuming it for review. Once the roof model passes a review, the application may then proceed with converting the roof model into a roof report.

Abstract: A computing device for dental impression scan merging includes a processor configured to generate a first model and a second model including a first and second plurality of geometric faces indicative of a first and second dental arch of a user. The processor generates a first point cloud of the first model and a second point cloud of the second model. The processor aligns the first point cloud and the second point cloud. The processor merges the first and second model to generate a merged model where merging the first and second model is based on the alignment of the first point cloud and the second point cloud.

Abstract: A user device can receive boundary information that identifies a restricted region; identify one or more map boundaries of the restricted region based on the boundary information; identify traversing roads of the restricted region based on geographic information survey (GIS) information; flag the traversing roads for exclusion from route determination; receive a navigation request identifying at least one of an origin or a destination; determine a route that does not include a flagged traversing road; and provide navigation information identifying the route.

Abstract: A tessellation method uses both vertex tessellation factors and displacement factors defined for each vertex of a patch, which may be a quad, a triangle or an isoline. The method is implemented in a computer graphics system and involves calculating a vertex tessellation factor for each corner vertex in one or more input patches. Tessellation is then performed on the plurality of input patches using the vertex tessellation factors. The tessellation operation involves adding one or more new vertices and calculating a displacement factor for each newly added vertex. A world space parameter for each vertex is subsequently determined by calculating a target world space parameter for each vertex and then modifying the target world space parameter for a vertex using the displacement factor for that vertex.

Abstract: Techniques of refining a model of a complex surface can include refining, upon insertion of a new control point, only those blending functions corresponding to the control points adjacent to the control point being inserted. The resulting curves cannot be considered as B-splines because the relationship between blending functions corresponding to adjacent control points no longer has a fixed relationship. Rather, the resulting blending functions, denoted herein as “S-splines,” sacrifice this fixed relationship in exchange for the ability to specify how many blending functions corresponding to adjacent control points are refined when performing exact local refinement. S-splines do not produce excess control points and have a simple representation. While the resulting functions representing the surface are not B-splines, they allow for exact refinement without affecting the blending functions corresponding to control points not neighboring the new control point.

Abstract: An exemplary virtual reality provider system generates a surface data projection based on surface data representative of color characteristics or depth characteristics of surfaces within a three-dimensional (“3D”) space. The surface data projection is generated for a portion of the 3D space. The virtual reality provider system applies an image transform operation to the surface data projection to transform the surface data projection. Specifically, the image transform operation transforms the surface data projection to account for a level of detail of the surfaces within the 3D space with respect to a particular vantage point within the 3D space. Corresponding methods and systems are also disclosed.

Abstract: An encoder is configured to compress point cloud geometry information using an octree geometric compression technique that utilizes a binary arithmetic encoder, a look-ahead table, a cache, and a context selection process, wherein encoding contexts are selected based, at least in part, on neighborhood configurations. In a similar manner, a decoder is configured to decode compressed point cloud geometry information utilizing a binary arithmetic encoder, a look-ahead table, a cache, and a context selection process.

Abstract: The disclosure notably relates to a computer-implemented method for designing a three-dimensional (3D) finite element mesh of a 3D part that includes a lattice structure. The method includes superposing a regular tiling of cells with a solid representation of the 3D part, partitioning the cells into two groups, a first group of cells, each in contact with the solid representation, and a second group of cells, none in contact with the solid representation. The method also includes computing a Boolean union of the first group of cells and the solid representation, the Boolean union forming a volume, finite element meshing the volume of the computed Boolean union while preserving the set of faces of the first group of cells that are shared with the second group of cells, and merging the finite element meshes of the cells of the second group and the meshed volume of the computed Boolean union.

Abstract: Ray tracing, and more generally, graphics operations taking place in a 3-D scene, involve a plurality of constituent graphics operations. Responsibility for executing these operations can be distributed among different sets of computation units. The sets of computation units each can execute a set of instructions on a parallelized set of input data elements and produce results. These results can be that the data elements can be categorized into different subsets, where each subset requires different processing as a next step. The data elements of these different subsets can be coalesced so that they are contiguous in a results set. The results set can be used to schedule additional computation, and if there are empty locations of a scheduling vector (after accounting for the members of a given subset), then those empty locations can be filled with other data elements that require the same further processing as that subset.

Abstract: A tessellation method uses tessellation factors defined for each vertex of a patch which may be a quad, a triangle or an isoline. The method is implemented in a computer graphics system and involves comparing the vertex tessellation factors to a threshold. If the vertex tessellation factors for either a left vertex or a right vertex, which define an edge of an initial patch, exceed the threshold, the edge is sub-divided by the addition of a new vertex which divides the edge into two parts and two new patches are formed. New vertex tessellation factors are calculated for each vertex in each of the newly formed patches, both of which include the newly added vertex. The method is then repeated for each of the newly formed patches until none of the vertex tessellation factors exceed the threshold.

Abstract: A user interface is adaptive to multiple display screens with different characteristics. An interface element encoded for a baseline display screen can be automatically transformed into a form suitable for a target display screen prior to rendering. Subsequently, a transformed encoded description of the interface element can be rendered in a displayable form such as an image and transmitted to a target display screen for display.

Abstract: An online analytical processing system may comprise an n-dimensional cube partitioned into slices, in which each slice may represent data points at the intersections of fixed and variable dimensions. Computation of data points within a slice may be deferred. A dependency graph may be initially constructed, in which the dependency graph is utilized in a subsequent computation. Calculation of data points may be prioritized based on information indicative of a chance that the data points will be accessed.

Abstract: Provided are a graphics processing unit and a graphics processing method for performing path rendering. The graphics processing method may include receiving object information including primitive information regarding the object; generating a primitive mask with respect to respective pixels, the primitive mask corresponding to the primitive information and including a plurality of bits; generating, based on the primitive mask, winding numbers with respect to the respective pixels; and rendering the pixels, based on the winding numbers.

Abstract: Autonomous vehicles and techniques that can be utilized to compress point cloud data and operate on compressed point cloud data are provided. An autonomous vehicle can include a data compression system can configure point cloud data according to a collection of three-dimensional (3D) tiles representative of the region. Each 3D tile can include a portion of the cloud point data, where each point vector in the portion of the cloud point data can be configured relative to a position vector of the 3D tile defined in a coordinate system of the collection of 3D tiles. The data compression system can utilize a fixed-point Q-format representation based on a defined number of bits to compress at least a portion of the point cloud data. The autonomous vehicle also can include a control system that can operate mathematically on compressed point cloud data, without reliance on prior decompression.

Abstract: An imaging system for a vehicle includes a rear backup camera and a video display screen for displaying video images derived from image data captured by the rear backup camera. The imaging system generates a backup overlay and an alignment overlay that are electronically superimposed on the displayed video images to assist a driver of the vehicle when executing a backup maneuver. The backup overlay includes a pair of side overlays superimposed on the displayed video images and the alignment overlay is superimposed on the displayed video images between the side overlays. The video display screen displays the video images and the backup overlay is generated responsive to the vehicle being shifted into reverse gear. The alignment overlay is generated responsive to a user input.

Abstract: Using various embodiments, methods and systems for displaying digital smart objects in 3D environments are described. In one embodiment, a system receives a request to present the 3D digital smart object in a game development environment of a game engine. The system can be configured to retrieve 3D digital smart object data from an asset repository, transmit the 3D digital smart object data to the game development environment of the game engine, receive a position location for the 3D digital smart object in the game, receive scaling information related to the 3D digital smart object, and store, into the asset repository, the position location, and scaling information related to the 3D digital smart object displayed in the game. Thereafter, the 3D digital smart object can be displayed at the position location when a player is interacting with the game at the game scene.

Abstract: In this disclosure, a solution for denoising a curve mesh is proposed. For a curve mesh including a polygonal facet, a noisy normal and a ground-truth normal of a first facet in the mesh is obtained. Then, based on the noisy normal, a first geometric feature of the first facet is determined from a plurality of neighboring facets of the first facet in the mesh. Next, based on the first geometric feature and the ground-truth normal, a mapping from the first geometric feature to the ground-truth normal of the first facet is determined for denoising the mesh.

Abstract: The disclosure notably relates to a computer-implemented method for designing a three-dimensional (3D) finite element mesh of a 3D part that comprises a lattice structure. The method includes superposing a regular tiling of cells with the solid representation of a 3D part, partitioning the cells into two groups, a first group of cells, each in contact with the solid representation of the 3D part, and a second group of cells, none in contact with the solid representation. The method also includes finite element meshing a boundary of the solid representation, extracting a boundary finite element mesh of the first group of cells, computing a Boolean union of the finite element mesh and the extracted boundary finite element mesh, finite element meshing a volume of the computed Boolean union and merging the finite element meshes of meshed volume of computed Boolean union and the cells of the second group of cells.

Abstract: In tile-based graphics processing systems, a tiling unit determines which tiles of a rendering space a primitive is in, such that the primitives in a tile can be rendered. Rather than performing tiling calculations for each tile in a bounding box for a primitive, tiling tests can be performed for a subset of the tiles. Then the results of the tiling tests for the subset of tiles can be used to determine whether the primitive is in other tiles which are located within a region bounded by two or more of the tiles of the subset. In this way the tiling process can be implemented without performing tiling calculations for all of the tiles in the bounding box for a primitive. Reducing the number of tiling calculations can help to improve the efficiency of the graphics processing system (in terms of speed and power consumption) for rendering a primitive.

Abstract: An encoding device includes an extracting unit that extracts, from voxel data expressing a three-dimensional shape by a collection of voxels arranged at lattice positions in a space, (i) structure information indicating presence/absence of a voxel at each lattice position and (ii) attribute information including an attribute value of each voxel; a structure encoding unit that encodes the structure information by determining a structure code corresponding to the structure information on a structure element having at least lattice position according to first context information indicating presence/absence of voxel at at least one predetermined lattice position around the structure element; and an attribute encoding unit that encodes an attribute value of a voxel of interest into an attribute code using second context information indicating presence/absence of a voxel at at least one predetermined lattice position around the voxel of interest.

Abstract: In implementations of freeform gradient style transfer, a style transfer system transfers a style of a freeform gradient of a source geometry of an object to a target geometry of an additional object. The system adds source mesh handles to a boundary of a source mesh of the object, and the system also adds target mesh handles to a boundary of a target mesh of the additional object. The source mesh handles are mapped to the target mesh handles, and the system warps the source mesh to appear as the target mesh based on the mapping. Color point locations of the freeform gradient are mapped from the source mesh to the warped source mesh to transfer the freeform gradient style from the source geometry of the object to the target geometry of the additional object.

Abstract: A tessellation method uses vertex tessellation factors. For a quad patch, the method involves comparing the vertex tessellation factors for each vertex of the quad patch to a threshold value and if none exceed the threshold, the quad is sub-divided into two or four triangles. If at least one of the four vertex tessellation factors exceeds the threshold, a recursive or iterative method is used which considers each vertex of the quad patch and determines how to further tessellate the patch dependent upon the value of the vertex tessellation factor of the selected vertex or dependent upon values of the vertex tessellation factors of the selected vertex and a neighbor vertex. A similar method is described for a triangle patch.

Abstract: A one-time write, read-only memory for storing trimming parameters includes an address pointer table, a fixed packet portion, and a flexible packet portion. The fixed packet portion includes one or more fixed packets, each fixed packet including trimming parameters for a component identified for trimming during a design phase. The flexible packet portion includes one or more flexible packets of different types. Each flexible packet includes trimming parameters for a component identified for trimming after the design phase. One packet type includes a length section and a number of fields equal to a value stored in the length section. Each field includes an address, a trimming parameter, and a mask. Another packet type includes trimming parameters associated with operands in operating instructions for a microcontroller, where the operands include an address and a mask.

Abstract: A smart device receives applications from locales wherein the application is useful only within the locale or a designated portion of the locale. The smart device includes a processor. The processor receives a signal indicating an application useful in a locale or a portion of the locale is available from a designated location, determines whether the device is authorized to download the application from the designated location, and if not, not download the application. If the device is authorized to download the application from the designated location, the processor determines whether the application is already stored in the device memory, and if so, runs the application when the smart device is in the designated portion of the locale, and if not, download the application from the designated location, and run the application when the smart device is in the locale or the designated portion of the locale.

Abstract: Functionality is disclosed herein for using a framework for a VR/AR application to utilize different services. In some configurations, a VR/AR application can utilize different services, such as an animation service, a multi-modal disambiguation service, a virtual platform service, a recognition service, an automatic speech recognition (ASR) service, a text-to-speech (TTS) service, a search service, as well as one or more other services. Instead of a developer of the VR/AR application having to develop programming code to implement features provided by one or more of services, the developer may utilize functionality of existing services that are available from a service provider network.

Abstract: This disclosure is directed to calibrating sensors mounted on an autonomous vehicle. First image data and second image data representing an environment can be captured by first and second cameras, respectively (and or a single camera at different points in time). Point pairs comprising a first point in the first image data and a second point in the second image data can be determined and projection errors associated with the points can be determined. A subset of point pairs can be determined, e.g., by excluding point pairs with the highest projection error. Calibration data associated with the subset of points can be determined and used to calibrate the cameras without the need for calibration infrastructure.

Abstract: A method and image processing apparatus for creating simplified representations of an existing virtual 3D model for use in occlusion culling is provided. A visual hull construction is performed on the existing virtual 3D model using an approximate voxel volume consisting of a plurality of voxels. A set of projections from a plurality of viewing angles provide a visual hull of the existing 3D model. The volumetric size of the visual hull of the existing 3D model is increased to envelop the existing virtual 3D model to provide the visual hull as an occludee model, and the volumetric size of the visual hull of the existing 3D model is decreased to be enveloped by the existing virtual 3D model to provide the visual hull as an occluder model. The occludee model and the occluder model are used during runtime in a 3D virtual environment for occlusion culling.

Abstract: A modeling apparatus calculates, with respect to edges in a three-dimensional mesh model, an evaluation value of each edge on the basis of the length of the edge, and selects the edges one by one as a first edge in ascending order of their evaluation value. In this connection, a lower evaluation value is calculated for a shorter edge. Then, if there is a specific node placement position where a value of a calculation formula using distances between the node placement position and each first face having either one of first nodes at both ends of the selected first edge as a vertex is lower than or equal to a threshold, the modeling apparatus removes the first edge, first nodes, and first faces from the three-dimensional mesh model, and adds a second node at the specific node placement position in the three-dimensional mesh model.

Abstract: A method of rendering at least one of paths forming an object includes setting an initial value to each of tiles included in a frame based on a position and a proceeding direction of the at least one of paths, calculating a winding number of each of the tiles through which the at least one of paths passes, among the tiles included in the frame, based on the set initial value, and determining whether to perform shading based on the set initial value and the calculated winding number.

Abstract: A method, system and computer program product for transmitting content across a link in a communications network. In an embodiment, the method comprises examining traffic in the network at a sender end of the link to determine if the traffic is carrying compressed content; when the traffic is carrying compressed content, decompressing at least one portion of the compressed content; and checking a cache to determine if the cache has content matching the at least one decompressed portion of the content. In an embodiment, the method further comprises when the cache has content matching the at least one decompressed portion of the content, determining an identifier associated with the at least one decompressed portion of the content; and using the identifier to reconstruct, at a receiving end of the link, the at least one decompressed portion of the content.

Abstract: A graphics processing pipeline includes: a position shader, a tiler, a pool of memory for storing primitive lists and vertex shaded attributes data for vertices, a varying-only vertex shader, and a fragment frontend and shader. The position shader performs vertex shading for the positional attributes of the vertices of a set of vertices to be processed by the graphics processing pipeline. The tiler uses the vertex shaded position data to identify primitives that should be processed further to generate the render output. When the tiler determines that a vertex should be processed further to generate the render output, it allocates memory space in the memory pool for storing vertex shaded attributes data for the vertex. Vertex shaded attributes data for the vertex is then stored in the allocated space in the memory pool for later use, e.g., by the fragment frontend and shader.

Abstract: The specification describes methods and systems for increasing a dimensional depth of a two-dimensional image of a face to yield a face image that appears three dimensional. The methods and systems identify key points on the 2-D image, obtain a texture map for the 2-D image, determines one or more proportions within the 2-D image, and adjusts the texture map of the 3-D model based on the determined one or more proportions within the 2-D image.

Abstract: Techniques of simplifying triangular meshes involve performing machine learning on triangle mesh data to produce simplified triangle mesh data representing a simplified triangle mesh having a specified number of triangles. Along these lines, a computer obtains triangular mesh data representing a triangular mesh that approximates an object. For example, a triangular mesh that provides a very good approximation of a human head may have tens or hundreds of thousands of triangular faces and vertices. The computer then inputs this triangular mesh data and a specified number of triangular faces into a machine learning application. The specified number of triangular faces is much less than the number of faces in the triangular mesh data input into the machine learning application.