# US 2025/0095172 A1 · Shape space generation via progressive correspondence estimation > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2025/0095172 A1 - Document ID: US20250095172A1 - Status: Published application - Kind: A1 - Application number: 18/369,958 - Filed: 2023-09-19 - Published: 2025-03-20 - Earliest claimed: 2023-09-19 - Assignees: Adobe Inc. - Applicants: Adobe Inc. - Inventors: Sanjeev Muralikrishnan; Chun-Hao Huang; Duygu Ceylan Aksit; Niloy J. Mitra - Web page: https://patentgrep.com/patent/US20250095172A1 ## CPC - G06T 7/38: Registration of image sequences - G06T 19/20: Editing of three-dimensional [3D] images, e.g. changing shapes or colours, aligning objects or positioning parts - G06T 7/344: {involving models} - G06T 7/75: {involving models} - G06T 17/10: Constructive solid geometry [CSG] using solid primitives, e.g. cylinders, cubes - G06T 2207/10028: Range image; Depth image; 3D point clouds - G06T 2207/20081: Training; Learning - G06T 2207/30196: Human being; Person - G06T 2219/2021: Shape modification - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract In some examples, a computing system access a set of registered three-dimensional (3D) digital shapes. The set of registered 3D digital shapes are registered to a shape template. The computing system determines a linear model for an estimate of the shape space using a first subset of the set of registered 3D digital shapes. The computing system then determines a nonlinear deformation model for the shape space using a second subset of the set of registered 3D digital shapes. An unregistered shape can be registered to the shape space using the linear model and the nonlinear deformation model. The registration can be added to the set of registered 3D digital shapes to update the estimate of the shape space if a shape distance between the registration and the unregistered shape is below a threshold value. ## Claims ### Claim 1 A method performed by one or more processing devices, comprising: accessing a set of registered three-dimensional (3D) digital shapes, wherein the set of registered 3D digital shapes are registered to a shape template; determining a linear model for an estimate of a shape space using a first subset of the set of registered 3D digital shapes; determining a nonlinear deformation model for the shape space using a second subset of the set of registered 3D digital shapes; projecting an unregistered shape to the shape space by using the linear model to create an initial registration for the unregistered shape; predicting an updated registration based on the initial registration using the trained nonlinear deformation model; and adding the updated registration to the set of registered 3D digital shapes based on a shape distance between the updated registration and the unregistered shape being below a threshold value for updating the estimate of the shape space to create an updated estimate of the shape space, wherein the updated estimate of the shape space is usable for editing a 3D digital shape in the shape space. ### Claim 2 (depends on 1) The method of claim 1, wherein the shape template is a skinned multi-person linear model (SMPL)-based template, wherein the linear model is a principal component analysis (PCA)-based model, comprising multiple shape eigenvectors, and wherein the nonlinear deformation model is a Neural Jacobian Fields (NJF)-based model. ### Claim 3 (depends on 1) The method of claim 1, wherein an initial state of the nonlinear deformation model is determined based on the first subset of the set of registered 3D digital shapes. ### Claim 4 (depends on 1) The method of claim 1, wherein projecting the unregistered shape to the shape space by using the linear model to create an initial registration for the unregistered shape further comprises: determining a plurality of optimized pose parameters and a plurality of optimized shape coefficients for the unregistered shape; and creating the initial registration for the unregistered shape based on the plurality of optimized shape coefficients and the linear model. ### Claim 5 (depends on 4) The method of claim 4, wherein the updated registration is posed to match the unregistered shape based on the plurality of optimized pose parameters by using the nonlinear deformation model. ### Claim 6 (depends on 1) The method of claim 1, wherein the shape distance is a Chamfer Distance. ### Claim 7 (depends on 1) The method of claim 1, further comprising: adding the updated registration to the first subset of the set of registered 3D digital shapes to create an updated first subset of the set of registered 3D digital shapes; determining an updated linear model for the shape space using the updated first subset of the set of registered 3D digital shapes; updating the nonlinear deformation model for the shape space using the second subset of the set of registered 3D digital shapes to create an updated nonlinear deformation model; projecting a second unregistered shape to the shape space by using the updated linear model to create a second initial registration; predicting a second updated registration based on the second initial registration using the updated nonlinear deformation model; determining a second shape distance between the second updated registration and the second unregistered shape; and adding the second updated registration to the updated first subset of the set of registered 3D digital shapes based on a second shape distance between the second updated registration and the second unregistered shape being below the threshold value to create a further updated first set of registered 3D digital shapes and a further updated estimate of the shape space. ### Claim 8 A system, comprising: a memory component; a processing device coupled to the memory component, the processing device to perform operations comprising: accessing a set of registered three-dimensional (3D) digital shapes, wherein the set of registered 3D digital shapes are registered to a shape template; determining a linear model for an estimate of a shape space using a first subset of the set of registered 3D digital shapes; determining a nonlinear deformation model for the shape space using a second subset of the set of registered 3D digital shapes; projecting an unregistered shape to the shape space by using the linear model to create an initial registration for the unregistered shape; predicting an updated registration based on the initial registration using the nonlinear deformation model; and adding the updated registration to the first subset of the set of registered 3D digital shapes based on a shape distance between the updated registration and the unregistered shape being below a threshold value to create an updated first subset of the set of registered 3D digital shapes for updating the estimate of the shape space to create an updated estimate of the shape space, wherein the updated estimate of the shape space is usable for editing a 3D digital shape in the shape space. ### Claim 9 (depends on 8) The system of claim 8, wherein the shape template is a skinned multi-person linear model (SMPL)-based template, wherein the linear model is a principal component analysis (PCA)-based model, comprising multiple shape eigenvectors, and wherein the nonlinear deformation model is a Neural Jacobian Fields (NJF)-based model. ### Claim 10 (depends on 8) The system of claim 8, wherein an initial state of the nonlinear deformation model is determined based on the first subset of the set of registered 3D digital shapes. ### Claim 11 (depends on 8) The system of claim 8, wherein projecting the unregistered shape to the shape space by using the linear model further comprises: determining a plurality of optimized pose parameters and a plurality of optimized shape coefficients for the unregistered shape; and identify a registered shape from the first subset of the set of registered 3D digital shapes best matching the unregistered shape based on the plurality of optimized shape coefficients. ### Claim 12 (depends on 11) The system of claim 11, wherein the updated registration is posed to match the unregistered shape based on the plurality of optimized pose parameters by using the nonlinear deformation model. ### Claim 13 (depends on 8) The system of claim 8, wherein the shape distance is a Chamfer Distance. ### Claim 14 (depends on 8) The system of claim 8, wherein the processing device is to perform further operations comprising: determining an updated linear model for the shape space using the updated first subset of the set of registered 3D digital shapes; updating the nonlinear deformation model for the shape space using the second subset of the set of registered 3D digital shapes to create an updated nonlinear deformation model; projecting a second unregistered shape to the shape space by using the updated linear model to create a second initial registration; predicting a second updated registration based on the second initial registration using the updated nonlinear deformation model; determining a second shape distance between the second updated registration and the second unregistered shape; and adding the second updated registration to the updated first subset of the set of registered 3D digital shapes based on a second shape distance between the second updated registration and the second unregistered shape being below the threshold value to create a further updated first set of registered 3D digital shapes and a further updated estimate of the shape space. ### Claim 15 A non-transitory computer-readable medium, storing executable instructions, which when executed by a processing device, cause the processing device to perform operations comprising: accessing a set of registered three-dimensional (3D) digital shapes, wherein the set of registered 3D digital shapes are registered to a shape template; a step for determining a linear model for an estimate of a shape space using a first subset of the set of registered 3D digital shapes; a step for obtaining a nonlinear deformation model for the shape space using a second subset of the set of registered 3D digital shapes; projecting an unregistered shape to the shape space by using the linear model to create an initial registration for the unregistered shape; predicting an updated registration based on the initial registration using the nonlinear deformation model; and adding the updated registration to the set of registered 3D digital shapes based on a shape distance between the updated registration and the unregistered shape being below a threshold value for updating the estimate of the shape space to create an updated estimate of the shape space, wherein the updated estimate of the shape space is usable for editing a 3D digital shape. ### Claim 16 (depends on 15) The non-transitory computer-readable medium of claim 15, wherein the shape template is a skinned multi-person linear model (SMPL)-based template, wherein the linear model is a principal component analysis (PCA)-based model, comprising multiple shape eigenvectors, and wherein the nonlinear deformation model is a Neural Jacobian Fields (NJF)-based model. ### Claim 17 (depends on 15) The non-transitory computer-readable medium of claim 15, wherein an initial state of the nonlinear deformation model is determined based on the first subset of the set of registered 3D digital shapes. ### Claim 18 (depends on 15) The non-transitory computer-readable medium of claim 15, wherein projecting the unregistered shape to the shape space by using the linear model further comprises: determining a plurality of optimized pose parameters and a plurality of optimized shape coefficients for the unregistered shape; and identify a registered shape from the first subset of the set of registered 3D digital shapes best matching the unregistered shape based on the plurality of optimized shape coefficients. ### Claim 19 (depends on 15) The non-transitory computer-readable medium of claim 15, wherein the updated registration is posed to match the unregistered shape based on the plurality of optimized pose parameters by using the nonlinear deformation model. ### Claim 20 (depends on 15) The non-transitory computer-readable medium of claim 15, wherein the executable instructions, which when executed by a processing device, cause the processing device to perform further operations comprising: adding the updated registration to the first subset of the set of registered 3D digital shapes to create an updated first subset of the set of registered 3D digital shapes; determining an updated linear model for the shape space using the updated first subset of the set of registered 3D digital shapes; updating the trained nonlinear deformation model for the shape space using the second subset of the set of registered 3D digital shapes to create an updated nonlinear deformation model; projecting a second unregistered shape to the shape space by using the updated linear model to create a second initial registration; predicting a second updated registration based on the second initial registration using the updated nonlinear deformation model; determining a second shape distance between the second updated registration and the second unregistered shape; and adding the second updated registration to the updated first subset of the set of registered 3D digital shapes based on a second shape distance between the second updated registration and the second unregistered shape being below the threshold value to create a further updated first set of registered 3D digital shapes and a further updated estimate of the shape space.