Multi-object vector sketch animation generation method based on grouping and motion trail prior

Abstract

The invention discloses a multi-object vector sketch animation generation method based on grouping and motion trail prior. According to the method, a two-stage process including'motion initialization 'and'motion optimization' is adopted. In a first stage, through an interactive canvas, a user can semantically group input static vector sketches, define a key frame motion trail for each group, and then generate a rough animation sequence through interpolation. In a second stage, the coarse animation sequence is input into a novel, group-based displacement network. The network independently predicts a displacement field for each semantic group to optimize motion and uses priori knowledge of a pre-trained text-to-video model for guidance and optimization. According to the method, manual priori of grouping and motion trails is introduced, so that motion among multiple objects is effectively decoupled, the animation quality and time sequence consistency are remarkably improved, and the method is particularly suitable for complex scenes containing multi-object interaction.

Description

Technical Field

[0001] This invention relates to the fields of computer graphics and artificial intelligence, and more specifically to a method for generating multi-object vector sketch animations based on grouping and motion trajectory priors. It is a text-guided technique for generating multi-object interactive vector sketch animations, applicable to applications such as animation production, game design, and education. Background Technology

[0002] Currently, with the development of generative models, text-driven sketch animation generation technology has made significant progress. Most existing methods utilize the motion prior knowledge contained in pre-trained text-to-video (T2V) diffusion models to impart dynamic effects to static sketches through iterative optimization.

[0003] However, existing technologies still have significant shortcomings when handling complex scenes involving multiple interactive objects. On the one hand, models designed for single objects struggle to decouple the motion relationships between different objects when applied to multi-object scenes, often leading to objects in the animation exhibiting a "sticking" phenomenon that violates physical laws. On the other hand, while raster image-based methods can generate scene-level animations, they struggle to guarantee the stability and consistency of object appearance over time and have limited generalization capabilities. These limitations hinder the application of sketch animation technology in real-world scenarios.

[0004] Therefore, developing a method to generate high-quality, temporally coherent, and physically reasonable multi-object interactive animations has significant theoretical and practical value. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this invention provides a vector sketch animation generation method that can effectively handle multi-object interactions and complex motions. By introducing semantic grouping and motion trajectory priors, and combined with the guidance of a pre-trained T2V model, it can ultimately generate high-quality and semantically accurate multi-object animations.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] In a first aspect, the present invention provides a method for generating multi-object vector sketch animations based on grouping and prior knowledge of motion trajectories. This method includes a two-stage process of motion initialization and motion optimization, wherein:

[0008] In the motion initialization phase: a static sketch composed of multiple vector strokes is received, and the vector strokes are divided into multiple semantic groups through an interactive user interface; a set of motion trajectories consisting of multiple keyframes is defined for each semantic group, and an initial animation sequence is generated for each semantic group by interpolating between the keyframes;

[0009] In the motion optimization phase: the initial animation sequence of all semantic groups is input into a group-based displacement network (GDN). The GDN independently predicts a displacement field for each semantic group to optimize the motion of the initial animation sequence and obtain a refined animation sequence. A pre-trained text-to-video diffusion model (T2V) is used to guide and optimize the parameters of the GDN based on text prompts describing the desired animation content.

[0010] Furthermore, during the motion initialization phase, interpolation is performed between the keyframes using a linear interpolation method, in the form of:

[0011]

[0012] t=(t i -t1) / (t2-t1)∈[0,1]

[0013] in, Indicates intermediate frame t i The location of the central control point and , where represents the position of the corresponding control point in two adjacent keyframes, and t represents the relative time position of the intermediate frame between the two keyframes.

[0014] Furthermore, in the motion optimization phase, the group-based displacement network comprises multiple parallel sub-networks, each configured to independently process a corresponding semantic group.

[0015] Furthermore, each of the sub-networks predicts the displacement field through a dual-path architecture, which includes a grouped local path and a grouped global path, wherein:

[0016] The grouped local path is used to predict the point-by-point displacement of each control point within the semantic group to simulate non-rigid deformation;

[0017] The grouped global path is used to predict an independent affine transformation for the semantic group in each frame of the animation sequence to simulate rigid motion.

[0018] Furthermore, each of the sub-networks includes a context-conditional feature enhancement module for injecting global contextual information into the vector feature representation of each control point before predicting the displacement field;

[0019] The contextual feature enhancement module specifically includes: a frame-aware position encoding unit and a motion context learning unit, wherein:

[0020] The frame-aware position encoding unit is used to encode the intra-frame spatial position and cross-frame temporal position of each control point;

[0021] The motion context learning unit is used to extract visual features from the image sequence rendered from the initial animation sequence and fuse them into the vector feature representation.

[0022] Furthermore, the optimization process employs the Score Distillation Sampling (SDS) loss function, which is as follows:

[0023]

[0024] in, φ represents the gradient of the SDS loss function with respect to parameter θ. During optimization, this gradient is calculated to update the parameters of the GDN, making the generated animation more consistent with the text description. θ represents the trainable parameters of the network being optimized. In this invention, θ is all the weights and biases of the group-based displacement network (GDN). g(θ) represents the generator network. In this invention, it is a composite function comprising the GDN network and a differentiable renderer. After inputting parameter θ, it generates the final output. x represents the final output generated by the generator network g(θ). In this invention, x is the rendered animation video frame sequence. φ represents the parameters of the pre-trained, fixed T2V diffusion model. These parameters remain unchanged during optimization because its knowledge is only borrowed, not trained. ∈ represents a noise vector randomly sampled from a standard normal distribution (Gaussian distribution). w(t) is the weighting function determined by the diffusion schedule. x t represents the image after noise addition, which is generated by mixing the original generated image x with random noise ∈ at the intensity corresponding to time step t; y represents the conditional information, i.e. the text prompts entered by the user, such as "a man is shooting a basketball"; This represents the Jacobian matrix of the generator output x relative to its parameter θ. This is a crucial part, as it allows the gradient of the loss to propagate back to the parameter θ of the GDN via the chain rule, thereby updating the network.

[0025] Through iterative refinement, the loss converges to generate an animation that is semantically aligned with the given text description.

[0026] Secondly, the present invention also provides a multi-object vector sketch animation generation system based on grouping and motion trajectory prior, applied to the above-mentioned method, for generating multi-object vector sketch animations. This system includes: a motion initialization module and a motion optimization module, wherein:

[0027] The motion initialization module is used to receive a static sketch composed of multiple vector strokes, divide the vector strokes into multiple semantic groups through an interactive user interface, define a set of motion trajectories composed of multiple keyframes for each semantic group, and generate an initial animation sequence for each semantic group by interpolating between the keyframes.

[0028] The motion optimization module is used to input the initial animation sequence of all semantic groups into a group-based displacement network, predict a displacement field independently for each semantic group, optimize the motion of the initial animation sequence, and obtain a refined animation sequence; and use a pre-trained text-to-video diffusion model to guide and optimize the parameters of the group-based displacement network based on text prompts describing the desired animation content.

[0029] Thirdly, embodiments of the present invention also provide an electronic device, including a processor and a memory, wherein the memory stores machine-executable instructions that can be executed by the processor, and the processor executes the machine-executable instructions to implement the above-described method for generating multi-object vector sketch animation based on grouping and motion trajectory priors.

[0030] As can be seen from the above technical solution, the present invention provides a method for generating multi-object vector sketch animation based on grouping and prior knowledge of motion trajectories, which has at least the following beneficial effects compared with the prior art:

[0031] This invention, by introducing user-defined semantic grouping and motion trajectory priors, combined with the guidance of a pre-trained text-to-video diffusion model, ultimately generates high-quality and semantically accurate multi-object animations. This invention effectively decouples the motion between multiple objects and improves the quality and temporal consistency of the animation, enabling the generation of high-quality, temporally coherent, and logically consistent multi-object interactive animations. It is particularly suitable for complex scenarios involving multi-object interactions and has significant theoretical and practical value.

[0032] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0033] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0036] Figure 1 This is a flowchart illustrating a multi-object vector sketch animation generation method based on grouping and motion trajectory priors provided by the present invention.

[0037] Figure 2 This is a schematic diagram of the architecture of the group-based displacement network (GDN) provided by the present invention.

[0038] Figure 3 This is a schematic diagram illustrating grouping and motion trajectory definition using an interactive canvas, as provided in an embodiment of the present invention.

[0039] Figure 4 This is a schematic diagram of the electronic device structure provided in an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0041] In the description of this invention, it should be noted that some processes described in this application specification and drawings include multiple operations that appear in a specific order. However, it should be clearly understood that these operations may be performed in any order or in parallel. Furthermore, various numbers are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0042] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0043] See Figure 1As shown, this embodiment of the invention provides a method for generating multi-object vector sketch animation based on grouping and prior knowledge of motion trajectories, which mainly includes the following two steps:

[0044] Step S100, Motion Initialization: For an input static vector sketch, an interactive interface allows the user to segment its strokes into multiple independent groups based on semantics, and define the initial motion trajectory of each group by setting keyframes. Subsequently, the system automatically generates an animation sequence with a rough dynamic effect by interpolating between the keyframes.

[0045] Step S200, Motion Optimization: The coarse animation sequence generated in the initialization phase is input into a group-based displacement network (GDN). This network independently predicts a fine displacement field for each semantic group to optimize its motion. Finally, using a pre-trained T2V model and user-provided text cues, the GDN is optimized end-to-end using fractional distillation sampling (SDS) loss, ensuring that the final generated animation follows the user-defined trajectory while conforming to the dynamic semantics of the text description.

[0046] This invention effectively decouples the motion between multiple objects by introducing grouping and prior motion trajectories, and significantly improves the quality and temporal consistency of animation, making it particularly suitable for complex scenarios involving multiple object interactions.

[0047] The specific embodiments of the method of the present invention will be described in detail below with reference to the accompanying drawings:

[0048] The multi-object sketch animation generation method proposed in this invention includes the following steps:

[0049] Step S100, motion initialization. (e.g.) Figure 1 and Figure 3 As shown, the goal of this step is to generate a rough animation using prior knowledge provided by the user.

[0050] First, receive a static vector sketch P. init Using an interactive canvas, users can select strokes in a sketch and assign them to different semantic groups. For example, in a "shooting" scene, the strokes for the player and the basketball can be divided into two separate groups.

[0051] Then, by dragging each semantic group on the canvas, its spatial position is set at different points in time (i.e., keyframes), thereby defining its macroscopic motion trajectory.

[0052] Finally, the system fills the intermediate frames with user-defined sparse keyframes using a linear interpolation algorithm, generating an animation sequence with an initial motion trajectory. Where M is the number of semantic groups. For any two adjacent keyframes t1 and t2, the intermediate frame t... i control point location Calculated using the following formula:

[0053]

[0054] t=(t i -t1) / (t2-t1)∈[0,1]

[0055] Where t is the relative time position of the intermediate frame between the two keyframes.

[0056] Step S200, Motion Optimization. The goal of this step is to refine the coarse animation using prior knowledge from neural networks and T2V models. Its core is a group-based displacement network (GDN), the detailed structure of which is as follows: Figure 2 As shown, it can be broken down into the following sub-steps:

[0057] Step S210: Group Feature Extraction. Extract the initial animation sequence... The inputs are fed into M parallel subnetworks of the GDN. Each subnetwork first passes the input low-dimensional coordinate information through a multilayer perceptron. (N is the number of points, K is the number of frames) Embedded into a high-dimensional feature space, denoted as F pos .

[0058] Step S220: Context-conditioned Feature Enhancement (CCFE). This part aims to inject global context and temporal dynamic information into the vector features of each control point to improve the coherence and realism of the animation. It contains two key units:

[0059] 1. Frame-aware Positional Encoding (FPE): This unit is feature F pos It adds spatiotemporal location information. It employs a dual encoding strategy, including spatial location encoding and temporal location encoding. Spatial location encoding uses standard trigonometric function location encoding. Temporal location encoding uses a multi-frequency time encoding mechanism, as follows:

[0060] First, generate a set of frequency-attenuated base signals:

[0061]

[0062] Among them, T max =d is the frequency attenuation factor, I is the dimension index vector, and d is the feature dimension. This design allows low dimensions to correspond to high-frequency components and high dimensions to correspond to low-frequency components.

[0063] Then, calculate the outer product of the time step vector t and the frequency vector freqs:

[0064]

[0065] Then, the sine and cosine functions are used to map it to the interval [-1, 1] to obtain the basic time code. After the encoding is expanded and reshaped, it is processed through an MLP network with layer normalization to obtain the final time encoding.

[0066] Finally, the spatial and temporal codes are fused:

[0067]

[0068] Among them, F time This represents the vector feature resulting from the fusion of spatial and temporal encodings.

[0069] 2. Motion Context Learning (MCL): This unit injects visual priors into vector features F through multimodal fusion. time The process includes the following steps:

[0070] ① Visual feature extraction:

[0071] Use a differentiable renderer to initialize the animation sequence. The image is rendered as a sequence. Then, a pre-trained visual encoder (such as DINOv2) is used to extract the visual features of each frame, resulting in a feature tensor. Where K is the number of frames, P is the number of blocks per frame, and C is the feature dimension.

[0072] ② Temporal attention modeling:

[0073] To capture inter-frame dependencies, visual features are analyzed. A temporal attention module is applied. This module computes multi-head self-attention along the temporal dimension for each spatial patchposition, and then aggregates the temporal information through mean pooling to obtain the aggregated temporal features.

[0074] ③ Spatial cross-attention fusion:

[0075] To achieve interaction between vector features and visual features, a spatial cross-attention mechanism is employed. The vector features S, after layer normalization... norm =LayerNorm(Ftime As a query, it represents the projected, expanded time features. Using the key and value as keys, multi-head attention is computed to obtain the attention output A.

[0076] ④ Feature enhancement and multi-layer stacking:

[0077] Feature enhancement is achieved through residual connections and a feed-forward network (FFN):

[0078] F fused =F time +A

[0079] F enhanced =F fused +FFN(LayerNorm(F fused ))

[0080] To achieve deep feature interaction, the above spatiotemporal cross-attention process is stacked into L layers (e.g., L=6), and the vector features are refined layer by layer to obtain the final enhanced features.

[0081] Step S230: Grouped displacement field prediction. Each sub-network predicts the final displacement based on the enhanced features obtained in step S220 through a dual-path structure.

[0082] ① Grouped Local Path: Using an MLP to predict the fine displacement of each control point Used to simulate the non-rigid deformation of an object.

[0083] ② Grouped Global Path: Use another MLP to predict an independent affine transformation matrix T for each group in each frame. k It is used to simulate rigid body motions such as translation and rotation.

[0084] Global displacement through The calculation is performed, where G represents the geometric coordinate information of a vector semantic group. The subscript i represents the index of the semantic group. In an animation scene, each object or part that can move independently is defined as a semantic group. For example, in a "shooting" scene, i=1 might represent the "player" group, and i=2 might represent the "basketball" group. The superscript k represents the frame index in the animation sequence. It indicates that this is the data for the k-th frame.

[0085] Ultimately, the total displacement of each group is the sum of the local and global displacements: in This represents the displacement of the local path. This represents the displacement of the global path.

[0086] Step S240: Joint Optimization. Merge all optimized animation sequences after displacement updates and render them again into final video frames using a differentiable renderer. Then, feed the rendered video and the user-inputted text prompts into a frozen, pre-trained T2V diffusion model. Calculate the semantic difference between the rendered video and the text prompts using fractional distillation sampling (SDS) loss, and backpropagate this loss to update all trainable parameters in the GDN in an end-to-end manner. This process is iterated until the generated animation is smooth, natural, and highly consistent with the text description.

[0087] As described in the above embodiments, those skilled in the art will understand that this invention proposes a method for generating multi-object vector sketch animations based on grouping and motion trajectory priors, aiming to solve the problems of existing technologies in handling multi-object interactions and complex motions. This method employs a two-stage process comprising "motion initialization" and "motion optimization." In the first stage, through an interactive canvas, the user can semantically group the input static vector sketch and define keyframe motion trajectories for each group, subsequently generating a rough animation sequence through interpolation. In the second stage, the rough animation sequence is input into a novel group-based displacement network. This network independently predicts displacement fields for each semantic group to optimize motion, and utilizes prior knowledge from a pre-trained text-to-video (T2V) model for guidance and optimization. By introducing artificial priors for grouping and motion trajectories, this invention effectively decouples the motion between multiple objects and significantly improves the quality and temporal consistency of the animation, making it particularly suitable for complex scenarios involving multi-object interactions.

[0088] Furthermore, this embodiment of the invention also provides a multi-object vector sketch animation generation system based on grouping and motion trajectory prior, applied to the multi-object vector sketch animation generation method based on grouping and motion trajectory prior described in the above embodiments, to realize multi-object vector sketch animation generation. The system includes: a motion initialization module and a motion optimization module, wherein:

[0089] The motion initialization module is used to receive a static sketch composed of multiple vector strokes, divide the vector strokes into multiple semantic groups through an interactive user interface, define a set of motion trajectories composed of multiple keyframes for each semantic group, and generate an initial animation sequence for each semantic group by interpolating between the keyframes.

[0090] The motion optimization module is used to input the initial animation sequence of all semantic groups into a group-based displacement network, predict a displacement field independently for each semantic group, optimize the motion of the initial animation sequence, and obtain a refined animation sequence; and use a pre-trained text-to-video diffusion model to guide and optimize the parameters of the group-based displacement network based on text prompts describing the desired animation content.

[0091] The multi-object vector sketch animation generation system based on grouping and motion trajectory prior provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in this embodiment can be referred to the corresponding content in the aforementioned method embodiment, and will not be repeated here.

[0092] Furthermore, refer to Figure 4 As shown, this embodiment of the invention also provides an electronic device that can execute the above-described methods or systems. The electronic device may include a processor 10, a memory 11, a communication bus 12, and a communication interface 13, and may also include a computer program stored in the memory 11 and executable on the processor 10.

[0093] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits packaged with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connecting various components of the entire electronic device through various interfaces and lines. It executes programs or modules stored in the memory 11 and calls data stored in the memory 11 to perform various functions of the electronic device and process data.

[0094] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, electronic devices, or computer program products, etc. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0095] It should be noted that the word "comprising" does not exclude the presence of components or steps not listed in the claims. The words "a" or "an" preceding a component do not exclude the presence of a plurality of such components. This invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer.

[0096] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0097] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating multi-object vector sketch animation based on grouping and prior knowledge of motion trajectories, characterized in that, This method includes a two-stage process: motion initialization and motion optimization, wherein: In the motion initialization phase: a static sketch composed of multiple vector strokes is received, and the vector strokes are divided into multiple semantic groups through an interactive user interface; a set of motion trajectories consisting of multiple keyframes is defined for each semantic group, and an initial animation sequence is generated for each semantic group by interpolating between the keyframes; In the motion optimization phase: the initial animation sequence of all semantic groups is input into a group-based displacement network, a displacement field is predicted independently for each semantic group, the motion of the initial animation sequence is optimized, and a refined animation sequence is obtained; and a pre-trained text-to-video diffusion model is used to guide and optimize the parameters of the group-based displacement network based on text prompts describing the desired animation content.

2. The method for generating multi-object vector sketch animation based on grouping and motion trajectory priors according to claim 1, characterized in that, During the motion initialization phase, interpolation is performed between the keyframes using a linear interpolation method, in the form of: t=(t i -t1) / (t2-t1)∈[0,1] in, Indicates intermediate frame t i The location of the central control point and , where represents the position of the corresponding control point in two adjacent keyframes, and t represents the relative time position of the intermediate frame between the two keyframes.

3. The method for generating multi-object vector sketch animation based on grouping and motion trajectory priors according to claim 1, characterized in that, In the motion optimization phase, the group-based displacement network comprises multiple parallel sub-networks, each configured to independently process a corresponding semantic group.

4. The method for generating multi-object vector sketch animation based on grouping and motion trajectory priors according to claim 3, characterized in that, Each of the sub-networks predicts the displacement field through a two-path architecture, which includes a grouped local path and a grouped global path, wherein: The grouped local path is used to predict the point-by-point displacement of each control point within the semantic group to simulate non-rigid deformation; The grouped global path is used to predict an independent affine transformation for the semantic group in each frame of the animation sequence to simulate rigid motion.

5. The method for generating multi-object vector sketch animation based on grouping and motion trajectory priors according to claim 4, characterized in that, Each of the sub-networks includes a context-conditional feature enhancement module for injecting global contextual information into the vector feature representation of each control point before predicting the displacement field; The contextual feature enhancement module specifically includes: a frame-aware position encoding unit and a motion context learning unit, wherein: The frame-aware position encoding unit is used to encode the intra-frame spatial position and cross-frame temporal position of each control point; The motion context learning unit is used to extract visual features from the image sequence rendered from the initial animation sequence and fuse them into the vector feature representation.

6. The method for generating multi-object vector sketch animation based on grouping and motion trajectory priors according to claim 1, characterized in that, The optimization process employs a fractional distillation sampling loss function to ensure that the final generated animation follows the set trajectory while conforming to the dynamic semantics of the text description. The formula for the fractional distillation sampling loss function is: in, φ represents the gradient of the SDS loss function with respect to the parameter θ; φ represents the parameters of the pre-trained text-to-video diffusion model; g(θ) represents the generator network; x represents the final output generated by the generator network; ∈ represents a noise vector randomly sampled from a standard normal distribution; w(t) is the weighting function determined by the diffusion schedule; x t represents the image with added noise at time step t; y represents conditional information. This represents the Jacobian matrix of the generator network output x relative to the parameter θ.

7. A multi-object vector sketch animation generation system based on grouping and motion trajectory priors, characterized in that, The system includes a motion initialization module and a motion optimization module. When applied, it executes a multi-object vector sketch animation generation method based on grouping and motion trajectory prior as described in any one of claims 1 to 6, thereby realizing the generation of multi-object vector sketch animation.

8. An electronic device, characterized in that, It includes a processor and a memory, the memory storing machine-executable instructions that can be executed by the processor, the processor executing the machine-executable instructions to implement a multi-object vector sketch animation generation method based on grouping and motion trajectory priors as described in any one of claims 1 to 6.

Images