Three-dimensional model generation method compatible with open surface, electronic equipment and storage medium

Abstract

The invention discloses a three-dimensional model generation method compatible with an open surface, electronic equipment and a storage medium, and the method comprises the steps: learning mixed implicit representation through training a neural network, expressing a basic shape of a target three-dimensional model through the SDF of the mixed implicit representation, and adjusting the basic shape through the UDF of the mixed implicit representation. And finally obtaining a target three-dimensional model with an open surface. According to the method, the open surface and the non-manifold geometry which cannot be directly represented by the SDF and the occupancy field can be represented, meanwhile, the high-quality surface can be extracted more easily, and the problem that direct extraction of the surface in the UDF is very difficult is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of three-dimensional model generation, and in particular to a three-dimensional model generation method compatible with open surfaces, an electronic device and a storage medium. Background Art

[0002] 3D model generation is a key research topic in computer graphics, computer vision, and artificial intelligence, with widespread applications in gaming, film, virtual reality, augmented reality, industrial design, medical imaging, and other fields. Generating high-quality, diverse, and controllable 3D models has long been a research hotspot in this field. 3D model generation based on implicit neural representations has become the most popular approach in recent years. Implicit neural representations represent the shape of an object using a neural network that takes the coordinates of a point in space as input and outputs certain attributes of that point (such as SDF and occupancy).

[0003] Traditional implicit representation methods, such as Signed Distance Field (SDF) and Occupancy Field, are primarily used to represent closed, watertight geometries. These methods assume that the model's interior and exterior are well-defined, and extract the surface using zero-level isosurfaces. However, for objects with open boundaries or non-manifold geometry, such as clothing and thin structures (such as paper and leaves), SDFs and occupancy fields cannot directly represent their topology. This is because open surfaces (which can be understood as single-layer surfaces) do not have a well-defined "interior" and "exterior," and the zero-level isosurface of SDFs cannot correctly extract these surfaces. While it is possible to approximate open surfaces as two very thin, closed surfaces, this introduces additional computational complexity and can lead to inaccurate results, such as the appearance of undesirable holes or artifacts. Unsigned Distance Field (UDF) is a simplified form of distance field that only records the distance from a point to the nearest surface, but does not include sign information. In theory, UDFs can represent surfaces of arbitrary topological structure, including open surfaces. However, extracting high-quality surfaces from UDFs is very challenging. Traditional isosurface extraction methods (such as MarchingCubes) encounter many problems when applied to UDFs. For example, the generated surface may contain many holes, noise, and artifacts. This is because the zero isosurface of a UDF is often discontinuous, and the gradient changes dramatically near non-manifold regions, making the isosurface extraction algorithm unstable. Some methods attempt to extract surfaces by calculating the gradient of the UDF, but this is often computationally expensive and very sensitive to noise.

[0004] SDFs and occupancy fields cannot directly represent open surfaces and non-manifold geometries as they rely on the distinction between inside and outside. For open surfaces without clear inside and outside boundaries, it is impossible to extract the correct topology through the zero-level set. Extracting high-quality surfaces from UDFs is very difficult because the zero-level sets of UDFs are usually discontinuous and have sharp gradient changes, causing a large number of holes, noises, and artifacts in traditional isosurface extraction algorithms (such as Marching Cubes). Existing generative models are difficult to precisely control the topology of the generative model. The relationship between the latent space of implicit representation and the topology is complex and difficult to interpret, making it difficult to control the topology by adjusting the latent vectors. Moreover, existing methods lack a unified framework for handling both closed surfaces and open surfaces. Most methods focus either on closed surfaces or on open surfaces, lacking flexibility. There are indeed techniques for handling open surfaces in existing methods, but there is a lack of a unified framework for handling both closed surfaces and open surfaces. Most methods focus either on closed surfaces or on open surfaces, lacking flexibility. Summary of the Invention

[0005] To solve the above problems, this application proposes a three-dimensional model generation method, an electronic device, and a storage medium compatible with open surfaces.

[0006] The method includes: A three-dimensional model generation method compatible with open surfaces, including learning a hybrid implicit representation through training a neural network, using the SDF of the hybrid implicit representation to express the basic shape of the target three-dimensional model, and using the UDF of the hybrid implicit representation to adjust the basic shape to finally obtain a target three-dimensional model with an open surface.

[0007] Based on the above method, further, the basic shape expressed by the SDF is a closed shape that is topologically homeomorphic to the target shape but has no open boundaries; the UDF is used to delete a specified area on the closed basic shape to generate a target shape with open boundaries.

[0008] Based on the above method, further, the process of using the UDF to delete a specified area on the closed basic shape includes setting a threshold Ru, regarding the area where the unsigned distance u to the target shape is less than Ru as a valid area to be retained, and regarding the area where u >= Ru as an invalid area to be deleted.

[0009] Based on the above method, further, in the process of learning the hybrid implicit representation by training a neural network, the trained neural network can be one of a multi-layer perceptron model (MLP), a convolutional neural network model (CNN) or a Transformer model; the hybrid implicit representation is in the form of (s,u)=Network(p;θ), where p=(x,y,z) is the input value of the neural network, representing the three-dimensional coordinates of the point in space; θ represents the parameters of the neural network Network; s is the SDF value, representing the signed distance from the point p in space to the target shape; u is the UDF value, representing the unsigned distance from the point p in space to the target shape.

[0010] Based on the above method, the neural network training process further includes: training the neural network through multiple three-dimensional models to form training data and combining with the loss function, wherein the three-dimensional models each include a surface point cloud P composed of a group of sampling points uniformly sampled from the surface of the three-dimensional model, the SDF value of each sampling point and the UDF value of each sampling point.

[0011] Based on the above method, further, the neural network also includes an encoder for supporting conditional generation, and the encoder is used to encode the input conditional information into a potential vector z.

[0012] Based on the above method, the neural network further includes a decoder for supporting conditional generation, which takes the spatial coordinates of point p and the potential vector z as input and outputs the SDF value s and UDF value u of the point.

[0013] Based on the above method, further, the decoder is a multi-layer perceptron model (MLP).

[0014] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute any one of the above methods through the computer program.

[0015] A computer-readable storage medium includes a stored program, wherein the program executes any one of the above methods when executed by a processor.

[0016] This method can represent open surfaces and non-manifold geometry that cannot be directly represented by SDFs and occupancy fields. It also makes it easier to extract high-quality surfaces, resolving the difficulty of directly extracting surfaces in UDFs. Compared to explicit representation methods such as polygon meshes, this method can represent objects with arbitrary topology. Explicit representation methods generally struggle to handle complex topologies and require a large amount of storage space to represent high-resolution models. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flowchart of a method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0019] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0020] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood broadly. For example, they can refer to fixed, removable, or integral connections; mechanical or electrical connections; direct connections, indirect connections through an intermediary, or internal connections between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention in specific contexts. In the following description, for purposes of illustration and not limitation, specific details such as particular system structures and technologies are provided to facilitate a thorough understanding of the embodiments of the present application. However, those skilled in the art will appreciate that the present application may be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail. It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of the described features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or combinations thereof. It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0021] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0022] The present invention will be described in further detail below with reference to the accompanying drawings.

[0023] like Figure 1 As shown, one embodiment of the present application trains a neural network to learn a hybrid implicit representation, uses the SDF of the hybrid implicit representation to express the basic shape of the target 3D model, and uses the UDF of the hybrid implicit representation to adjust the basic shape to ultimately obtain a target 3D model with open surfaces. Based on the hybrid implicit neural representation, the present invention can simultaneously represent closed and open surfaces and achieve precise control over the topological structure of the generated model.

[0024] The core idea of ​​the method in this embodiment is to use SDF to represent a closed basic shape that is homeomorphic to the target shape, and then use a UDF to "carve" this basic shape, locating and deleting the parts that need to be carved according to the requirements of the target shape, thereby generating a target shape with an open boundary.

[0025] The hybrid implicit representation of the present invention includes a signed distance field representation and an unsigned distance field representation, wherein the signed distance field is SignedDistanceField (SDF), s=SDF(x).

[0026] s represents the signed distance from a point p in space to the target shape. s<0 indicates that point p is inside the target shape. s>0 indicates that point p is outside the target shape. s=0 indicates that point p is on the surface of the target shape. The SDF is used to represent a closed, watertight base shape that is topologically similar (homeomorphic) to the target shape but has no open boundaries.

[0027] Unsigned distance field, UnsignedDistanceField(UDF): u=UDF(p) u represents the unsigned distance from a point p in space to the target shape. u is always non - negative. The UDF is used to "sculpt" the base shape represented by the SDF to generate the target shape with open boundaries. Specifically, by setting a threshold Ru (usually a small positive number, such as 0.1), the embodiments of this application regard the region where u < Ru as the "valid region" and the region where u >= Ru as the "invalid region" (the region to be "sculpted" away). The finally generated surface will be near the isosurface where u = Ru.

[0028] The core of this application is that the SDF is responsible for representing a closed basic shape, and the UDF is responsible for "sculpting" on the basis of the SDF to remove the unnecessary parts of the shape represented by the SDF, thereby creating a shape with open boundaries. By controlling the threshold Ru of the UDF, the degree of sculpting can be controlled, thus finely controlling the topology of the finally generated model.

[0029] Based on the above embodiments, the present invention is further improved. The neural network of this application is usually a multi - layer perceptron (MLP), but it can also be other types of neural networks, such as a convolutional neural network (CNN) or a Transformer.

[0030] Based on the above embodiments, the present invention is further improved. In this embodiment, the input of the neural network is the three - dimensional coordinates p=(x, y, z) of a point p in space, and the output is the SDF value s and the UDF value u of this point, with the expression form: (s, u)=Network(p;θ), where θ represents the parameters of the neural network Network.

[0031] Based on the above embodiments, the present invention is further improved. In this embodiment, in order to support conditional generation, for example, generating a three - dimensional model based on text or image, this application adopts an encoder - decoder structure (Encoder - Decoder Structure).

[0032] Among them, the encoder (Encoder): z = Encoder(condition;θ_enc), where θ_enc represents the parameters of the encoder. The encoder is used to encode the conditional information (such as text description, image, etc.) into a latent vector z. The encoder can be any type of neural network. For example, for text conditions, a recurrent neural network (RNN) or a Transformer can be used; for image conditions, a convolutional neural network (CNN) can be used.

[0033] The decoder (Decoder): (s,u)=Decoder(p,z;θ_dec) takes the spatial coordinate p and the potential vector z as input and outputs the SDF value s and UDF value u of the point. The decoder is usually an MLP model, and θ_dec represents the decoder parameters.

[0034] Based on the above embodiment, the present invention further improves upon it. In this embodiment, to enhance the model's expressiveness and efficiency, this application uses a vector set (VecSet) as a latent variable. VecSet represents the model's features as a set of vectors. Compared to traditional single vectors or three-dimensional grids, VecSet achieves a better balance between expressiveness and computational efficiency.

[0035] The VecSet encoding process involves encoding the input point cloud (or mesh) into a set of latent vectors, specifically: Input point cloud: P={p1,p2,...,pN}; Feature extraction: fi=MLP(pi), get the feature vector fi of each point; Set sampling: Sample K vectors from {f1,f2,...,fN} to obtain VecSet:Z={z1,z2,...,zK}; the above sampling method can use the farthest point sampling method (farthestpointsamplingFPS).

[0036] The VecSet decoding process includes using the VecSet as a condition to decode the SDF and UDF values, specifically including: Input space coordinates p, VecSetZ={z1,z2,...,zK}; Concatenate the spatial coordinates with each zi to obtain K fused feature vectors: [p,z1], [p,z2], ..., [p,zK]; Input each fused feature vector into the shared weight MLP to obtain K outputs: o1, o2, ..., oK; Aggregate the K outputs to get the final SDF and UDF values: (s,u)=Aggregation(o1,o2,...,oK).

[0037] Based on the above embodiment, the present invention further improves upon the above embodiment. In this embodiment, training a neural network to learn a hybrid implicit representation is a key step of the present application. The neural network is trained using 3D model data. During the training process, a surface point cloud (SurfacePointCloud) is extracted for each model. The neural network is trained based on the SDF and UDF values ​​of the surface point cloud. The surface point cloud is a set of points uniformly sampled from the model surface. For each sampled point, the SDF value from the point to the model surface is calculated. The SDF value calculation process includes, for closed surfaces, using existing SDF calculation methods. For open surfaces, they are first converted into a closed, watertight proxy model, and then the SDF value is calculated. The proxy model can be constructed using a variety of methods, such as dilation, which dilates the open surface a small distance in both directions along the normal direction to form a thin closed body. For example, boundary completion finds the boundaries of the open surface and then uses a triangulation algorithm (such as Delaunay triangulation) to connect the boundaries to form a closed surface. The method preferably used in this application also includes finding the boundary of the open surface, grouping the boundary points (boundary points belonging to the same hole are grouped together), calculating a plane for each group of boundary points, projecting the boundary points onto the plane, and then triangulating the projected points. Finally, these triangular facets are added to the original model to form a closed proxy model.

[0038] Based on the above embodiment, the present invention further improves upon this. In this embodiment, the process for calculating the SDF value from each sampling point to the model surface is further improved. The distance from the point to all triangular facets in the model can be directly calculated, and the minimum value is taken as the UDF value. To accelerate the calculation, a spatial partitioning data structure (such as an octree or kd-tree) can be used to accelerate the nearest neighbor search.

[0039] Based on the above embodiment, the present invention is further improved. In this embodiment, a loss function is also introduced to train the neural network. The loss function is mainly obtained through the SDF loss function and the UDF loss function.

[0040] Wherein the SDF loss function: L_SDF=BCE(s_pred,s_gt) s_pred is the SDF value predicted by the network.

[0041] s_gt is the true SDF value (from the training data).

[0042] BCE is BinaryCrossEntropyLoss.

[0043] UDF loss function: \(L_{UDF} = \|u_{pred} - u_{gt}\|_2\) \(u_{pred}\) is the UDF value predicted by the network.

[0044] \(u_{gt}\) is the true UDF value (from the training data).

[0045] \(\|.\|_2\) is the L2 norm.

[0046] According to the SDF loss function and the UDF loss function, the total loss function (TotalLoss) of this application is obtained: \(L=\lambda_{SDF}*L_{SDF}+\lambda_{UDF}*L_{UDF}\). \(\lambda_{SDF}\) and \(\lambda_{UDF}\) are weight coefficients used to balance the importance of the SDF loss and the UDF loss, and both are usually set to 1.

[0047] In the specific training process, the optimizer (Optimizer) of this application usually uses the Adam or AdamW optimizer; the initial learning rate of the learning rate (LearningRate) is usually set to \(1e - 3\) or \(1e - 4\), and the learning rate decay strategy (such as cosine annealing) is used to gradually reduce the learning rate; the batch size (BatchSize) is usually set to 1024 or 2048; the number of training epochs (Epochs) is usually trained for 1000 or more epochs until the loss function converges; to prevent gradient explosion, gradient clipping (GradientClipping) technology can be used; to prevent overfitting, weight decay (WeightDecay) technology can be used.

[0048] Based on the above embodiments, the present invention is further improved. In order to more effectively utilize the UDF information during training, the embodiments of this application can perform binarization on the true UDF value \(u_{gt}\): `u_{gt\_binary}=(u_{gt}<Ru).float()` If `u_{gt}<Ru`, then `u_{gt\_binary}=1` If `u_{gt}>=Ru`, then `u_{gt\_binary}=0` Then, the binarized UDF value is used to calculate the UDF loss: `L_{UDF}=\|u_{pred}-u_{gt\_binary}\|_2` The advantage of doing this is that it can transform the learning of UDF into a binary classification problem, simplifying the learning of UDF; it can more clearly indicate which regions of the network should be "sculpted" away.

[0049] Based on the above embodiment, the present invention is further improved. In this embodiment, in order to improve the training efficiency and the quality of the generated model, this embodiment adopts a hybrid sampling strategy, and the process includes: 1. Surface Sampling: uniformly sample a portion of points from the model surface.

[0050] 2. Space Sampling: Uniformly sample a portion of points in the space around the model.

[0051] 3. Near-Surface Sampling: Sample a portion of points near the model surface.

[0052] Specifically, for each surface sampling point, a new sampling point is generated by randomly offsetting it by a small distance in the direction of its normal. This increases the sampling density near the model surface and improves the model's ability to represent details. The ratio of the three sampling methods can be adjusted based on the specific dataset and task. For example, the ratio of the three sampling methods can be set to 1:1:1 or 2:1:1.

[0053] After training, the neural network of the present application can extract the surface of the target 3D model from the mixed implicit representation. The extraction process includes Initial Surface Extraction: First, use the MarchingCubes algorithm to extract an initial surface mesh from the SDF. Since the SDF represents a closed, watertight base shape, this step results in a mesh with no open boundaries. mesh_initial = MarchingCubes(SDF); Surface sculpting: UDF is then used to "sculpt" the initial surface mesh to generate the target shape with an open boundary. The steps include: 1) Traverse each triangle face of the initial mesh, and for each triangle face, obtain the coordinates v1, v2, and v3 of its three vertices; 2) For each vertex, query its UDF value: u1=UDF(v1), u2=UDF(v2), u3=UDF(v3); 3) Determine whether clipping is necessary. If u1, u2, and u3 are all less than a threshold value, Ru, the triangle is retained. If u1, u2, and u3 are all greater than or equal to Ru, the triangle is removed. If u1, u2, and u3 have values ​​both less than and greater than Ru, the triangle needs to be clipped.

[0054] 4) Clipping Triangle Face, the process includes: If one vertex is inside the "valid area" and two vertices are outside the "invalid area" (OneVertexInside,TwoVerticesOutside): Find the vertex v_in in the "valid area" and the two vertices v_out1 and v_out2 outside the "invalid area", calculate the intersection v_intersect1 of the edge between v_in and v_out1 and the u=Ru isosurface, calculate the intersection v_intersect2 of the edge between v_in and v_out2 and the u=Ru isosurface, and generate a new triangle patch with vertices v_in, v_intersect1, and v_intersect2. In the process of calculating the intersection point, linear interpolation can be used to calculate the intersection point. v_intersect1=v_in+(v_out1-v_in)*(Ru-u_in) / (u_out1-u_in) v_intersect2=v_in+(v_out2-v_in)*(Ru-u_in) / (u_out2-u_in).

[0055] If two vertices are inside the "valid area" and one vertex is outside the "invalid area" (TwoVerticesInside, OneVertexOutside): Find the vertex v_out outside the "invalid region" and the two vertices v_in1 and v_in2 inside the "valid region." Calculate the intersection point v_intersect1 of the edge between v_out and v_in1 with the u=Ru isosurface. Calculate the intersection point v_intersect2 of the edge between v_out and v_in2 with the u=Ru isosurface, generating two new triangular patches. The vertices of the first triangular patch are v_in1, v_intersect1, and v_intersect2. The vertices of the second triangular patch are v_in1, v_intersect2, and v_in2. The intersection point calculation method is the same as above.

[0056] 5) Repeat steps 1)-4) until all triangles of the initial mesh are processed.

[0057] Based on the above embodiment, the present invention is further improved. In this embodiment, in order to further improve the quality of the generated surface and eliminate the error introduced by the UDF threshold Ru, the embodiment of the present application performs boundary shrinkage on the cropped mesh. The process includes: Extract boundary edges: Find all edges that belong to only one triangle patch; these edges form the boundary of the cropped mesh. Calculate the UDF gradient at boundary points: For each boundary point, calculate the UDF gradient. Since the UDF gradient cannot be directly learned, numerical methods can be used to approximate the gradient. For example, the central difference method can be used: ∇UDF(p)≈(UDF(p+ε*n)-UDF(p-ε*n)) / (2ε), where n is the normal vector at the boundary point (which can be obtained by averaging the normal vectors of adjacent triangles) and ε is a small positive number. Move boundary points along the UDF gradient: Move each boundary point a small distance in the negative direction of the UDF gradient. The movement distance can be proportional to the magnitude of the UDF gradient or set to a fixed value: v_new=v_old-α*∇UDF(v_old), where v_old is the original boundary point, v_new is the moved boundary point, and α is the step size. Retriangulation: Retriangulate the moved boundary points to obtain a new mesh. You can use algorithms such as Delaunay triangulation.

[0058] Based on the above embodiment, the present invention further improves upon it. In this embodiment, after cropping and boundary shrinkage, some small holes or artifacts may still remain. Post-processing techniques are used to further improve the quality of the generated model. For example, Hole Filling uses mesh repair algorithms (such as MeshFix) to fill small holes. Smoothing uses mesh smoothing algorithms (such as LaplacianSmoothing) to smooth the surface. Remeshing uses mesh re-division algorithms (such as InstantMeshes) to improve mesh quality. Through the above steps, the quality of the generated target 3D model is further improved.

[0059] The hybrid implicit representation of this invention can control the topology of the generated model in various ways. By adjusting the UDF threshold Ru, the degree of "carving" can be controlled, thereby changing the model's topology. A larger Ru value results in more "carving," and the resulting model becomes more "open." A smaller Ru value results in less "carving," and the resulting model becomes more "closed." In the encoder-decoder architecture, the topology of the generated model can be controlled by adding topological information to the conditional input. For example, an additional text label can be used to indicate whether the model should be closed or open, or an additional vector can be used to represent the model's genus. Multiple UDFs can be used to control the topology of different parts of the model. For example, one UDF can be used to control whether the cuffs of a garment are open, and another to control whether the collar is open. The UDF threshold Ru can be set as a learnable parameter, allowing the network to automatically learn the optimal threshold. After the model is generated, the topology can be explicitly edited by manually editing the UDF field. For example, connecting two previously separate surfaces creates a "bridge" between the two surfaces, and the UDF value of the "bridge" area is set to be less than Ru. Disconnect: Disconnect a previously connected surface to create a "barrier" at the desired location and set the UDF value of the "barrier" area to be greater than or equal to Ru. Add / Delete Holes: Add or delete holes in the model by modifying the UDF value of the hole area.

[0060] The method of this application breaks through the limitations of implicit representation and realizes the representation of open surfaces and non-manifold geometry. Traditional implicit representation methods (such as SDF and occupancy field) are mainly used to represent closed, watertight geometries and have difficulty handling open surfaces and non-manifold geometry. The hybrid implicit representation method proposed in this invention breaks through this limitation by combining the advantages of SDF and UDF, and can simultaneously represent closed and open surfaces. This is because SDF is responsible for representing a basic shape (usually closed), while UDF is responsible for "sculpting" the SDF, removing unnecessary parts, and creating a shape with an open boundary. This representation method naturally supports various topological structures.

[0061] The method proposed in this application achieves high-quality surface extraction, avoiding the issue of zero isosurfaces in UDFs. Extracting high-quality surfaces from UDFs has always been a challenge, and traditional methods based on zero isosurfaces are prone to generating holes and artifacts. The proposed method, based on non-zero isosurfaces of UDFs, cleverly avoids the discontinuity issue of zero isosurfaces in UDFs by setting a threshold, Ru, resulting in smooth, continuous, and artifact-free surfaces. Furthermore, the proposed boundary contraction algorithm further improves the quality of the generated surfaces. The method of this application provides a flexible topology control mechanism, enabling the generation of 3D models with controllable topology. The present invention provides multiple methods for controlling the topology of generated models, including adjusting the UDF threshold Ru, using conditional inputs, multiple UDFs, learnable UDF thresholds, and explicit topology editing. These methods allow users to precisely control the topology of generated models as needed, for example, specifying whether the model is closed or open, or controlling the number and location of holes in the model.

[0062] The method of this application provides a unified framework that simplifies the generation of complex models: This invention provides a unified framework that can handle both closed and open surfaces, avoiding the complexity of training multiple models separately. This makes the generation of 3D models with complex topologies much simpler and more efficient. For example, when generating a clothed human model, only a single model needs to be trained to simultaneously generate the human body (closed surfaces) and clothing (open surfaces).

[0063] The proposed method is easy to integrate and highly scalable: The proposed method can be easily integrated into existing SDF-based generation models by simply adding an additional UDF output. Furthermore, the proposed method can be extended to various application scenarios, such as image-to-3D model generation, text-to-3D model generation, and component-based generation.

[0064] In summary, the method proposed in this paper is superior to existing technologies in terms of representation ability, surface extraction quality, topology control ability, generation process and scalability, providing a more powerful, flexible and easy-to-use tool for the field of 3D model generation.

[0065] The present application also discloses an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the method described in any one of the above embodiments through the computer program.

[0066] The present application also discloses a computer-readable storage medium, wherein the computer-readable storage medium includes a stored program, wherein the program is executed by a processor to execute the method described in any one of the embodiments. The present application implements all or part of the processes in the above-mentioned embodiment method by instructing the relevant hardware through a computer program. The computer program includes computer program code, which can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include at least: any entity or device capable of carrying computer program code to a device / electronic device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, mobile hard disk, magnetic disk or optical disk. In some jurisdictions, based on legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0067] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0068] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0069] In the embodiments provided in the present application, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device controller embodiments described above are merely schematic. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0070] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0071] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

[0072] The scope of protection of the present invention is not limited to this, and any changes or replacements of the technical solutions that can be thought of without creative work should be included in the scope of protection of the present invention. The scope of protection of the present invention shall be based on the scope of protection defined in the claims.

Claims

1. A method for generating a three-dimensional model compatible with open surfaces, characterized in that , including learning a hybrid implicit representation by training a neural network, using the signed distance function (SDF) of the hybrid implicit representation to represent the basic shape of the target 3D model, and using the unsigned distance function (UDF) of the hybrid implicit representation to adjust the basic shape to finally obtain the target 3D model with an open surface.

2. The method according to claim 1, characterized in that The basic shape represented by the SDF is a closed shape that is topologically homeomorphic to the target shape but has no open boundaries; the UDF is used to delete a specified area on the closed basic shape to generate a target shape with open boundaries.

3. The method according to claim 2, characterized in that The process of using the UDF to delete a specified area on the closed basic shape includes setting a threshold Ru, considering the area where the unsigned distance u to the target shape is u < Ru as a valid area to be retained, and considering the area where u >= Ru as an invalid area to be deleted.

4. The method according to claim 1, wherein In the process of learning the hybrid implicit representation by training the neural network, the trained neural network can be one of a multi-layer perceptron model (MLP), a convolutional neural network model (CNN), or a Transformer model; The hybrid implicit representation form is (s, u) = Network(p; θ), where p = (x, y, z) is the input value of the neural network, representing the three-dimensional coordinates of a point in space; θ represents the parameters of the neural network Network; s is the SDF value, representing the signed distance from the point p in space to the target shape; u is the UDF value, representing the unsigned distance from the point p in space to the target shape.

5. The method according to claim 4, characterized in that The neural network training process includes: training the neural network by using a training data set composed of multiple 3D models in combination with a loss function, where each 3D model includes a surface point cloud P composed of a set of sampled points uniformly sampled from the surface of the 3D model, the SDF value of each sampled point, and the UDF value of each sampled point.

6. The method according to claim 4, characterized in that The neural network further includes an encoder for supporting conditional generation, which is used to encode the input conditional information into a latent vector z.

7. The method according to claim 6, characterized in that The neural network further includes a decoder for supporting conditional generation, which takes the spatial coordinates of the point p and the latent vector z as inputs and outputs the SDF value s and the UDF value u of the point.

8. The method according to claim 7, characterized in that The decoder is a multi-layer perceptron model (MLP).

9. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method described in any one of claims 1 to 8 through the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, where the program, when run by the processor, executes the method described in any one of claims 1 to 9.

Images