CN111583422B - Heuristic editing method and device for three-dimensional human body model - Google Patents

Abstract

The invention discloses a heuristic editing method and a device of a three-dimensional human body model, wherein the method comprises the following steps: acquiring a three-dimensional human body model to be edited; converting a three-dimensional human body model to be edited into a geometric image by using a curved surface parameterization method; and the processing geometric image trains the two-dimensional neural network, outputs a generated geometric image, and inverts the generated geometric image into three-dimensional information. The method only needs to select the vertex without defining the sketch plane, thereby getting rid of the limitation of visual angles and simplifying the editing steps.

Description

Heuristic editing method and device for 3D human body model

technical field

The invention relates to the technical fields of computer vision and computer graphics, in particular to a heuristic editing method and device for a three-dimensional human body model.

Background technique

Editing of 3D human body models is an important problem in the field of computer graphics. However, the cost of editing 3D human body models is often very high. The traditional 3D model editing methods are cumbersome, require professional model animators to complete, and are time-consuming and labor-intensive.

In recent years, the rapid development of deep neural networks has provided a solution to many "from scratch" problems. However, the maturity of the 3D neural network is not as good as that of the 2D neural network. Therefore, the 3D problem cannot be solved by the method based on the neural network like the 2D problem.

Contents of the invention

The present invention aims to solve one of the technical problems in the related art at least to a certain extent.

Therefore, an object of the present invention is to propose a heuristic editing method for a three-dimensional human body model, which can greatly simplify the editing of the three-dimensional human body model.

Another object of the present invention is to propose a heuristic editing device for a three-dimensional human body model.

In order to achieve the above purpose, an embodiment of the present invention proposes a heuristic editing method for a three-dimensional human body model, including the following steps: step S1, obtaining the three-dimensional human body model to be edited; step S2, using the surface parameterization method to convert the The three-dimensional human body model is converted into a geometric image; step S3, processing the geometric image to train a two-dimensional neural network, outputting a generated geometric image, and inverting the generated geometric image into three-dimensional information.

The heuristic editing method of the 3D human body model in the embodiment of the present invention uses surface parameterization and deep neural network technology to artificially provide heuristic sketches to generate a human body model with corresponding details, which greatly simplifies the editing of the 3D human body model.

In addition, the heuristic editing method for a three-dimensional human body model according to the above-mentioned embodiments of the present invention may also have the following additional technical features:

Further, in one embodiment of the present invention, the step S3 includes: extracting the line image of the geometric figure through an edge detection algorithm, and randomly generating a mask image; performing a process on the geometric image and the mask image The mask is calculated to obtain a two-dimensional incomplete image; the two-dimensional neural network is trained by using the two-dimensional incomplete image.

Further, in an embodiment of the present invention, the step S1 also obtains the editing area selected by the user.

Further, in one embodiment of the present invention, before obtaining the geometric image, it is necessary to select one or more vertex chains to be deformed in the editing area selected by the user, and use the vertex chains to generate a line image, i.e. Edit Sketch.

Further, in an embodiment of the present invention, a mask image is mapped out according to the editing area selected by the user.

In order to achieve the above object, another embodiment of the present invention proposes a heuristic editing device for a 3D human body model, including: an acquisition module for acquiring a 3D human body model to be edited; a conversion module for converting the 3D human body model to The three-dimensional human body model to be edited is converted into a geometric image; the training module is used to process the geometric image to train the two-dimensional neural network, output the generated geometric image, and reversely convert the generated geometric image into three-dimensional information.

The heuristic editing device for the 3D human body model in the embodiment of the present invention uses surface parameterization and deep neural network technology to artificially provide heuristic sketches to generate a human body model with corresponding details, which greatly simplifies the editing of the 3D human body model.

In addition, the device for heuristic editing of a three-dimensional human body model according to the above-mentioned embodiments of the present invention may also have the following additional technical features:

Further, in one embodiment of the present invention, the training module includes: an extraction unit, used to extract the line image of the geometric figure through an edge detection algorithm, and randomly generate a mask image; performing mask calculation on the geometric image and the mask image to obtain a two-dimensional incomplete image; a training unit configured to use the two-dimensional incomplete image to train the two-dimensional neural network.

Further, in an embodiment of the present invention, the obtaining module is also used to obtain the editing area selected by the user.

Further, in one embodiment of the present invention, before obtaining the geometric image, it is necessary to select one or more vertex chains to be deformed in the editing area selected by the user, and use the vertex chains to generate a line image that is Edit the sketch.

Further, in an embodiment of the present invention, a mask image is mapped out according to the editing area selected by the user.

Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Description of drawings

The above and/or additional aspects and advantages of the present invention will become apparent and easy to understand from the following description of the embodiments in conjunction with the accompanying drawings, wherein:

Fig. 1 is a flowchart of a heuristic editing method for a three-dimensional human body model according to an embodiment of the present invention;

Fig. 2 is a schematic structural diagram of a heuristic editing device for a 3D human body model according to an embodiment of the present invention.

Detailed ways

Embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the figures are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

The following describes the heuristic editing method and device for a 3D human body model according to the embodiments of the present invention with reference to the accompanying drawings. First, the heuristic editing method for the 3D human body model according to the embodiments of the present invention will be described with reference to the accompanying drawings.

It should be noted that the so-called heuristic editing method refers to the editor providing a "suggestion" to the editor. The results generated by the editor may not be exactly the same as the "suggestions", but corresponding modifications will be made in the direction of the "suggestions". That is to say, the editing process is greatly simplified while sacrificing a certain editing accuracy.

Fig. 1 is a flowchart of a heuristic editing method for a 3D human body model according to an embodiment of the present invention.

As shown in Figure 1, the heuristic editing method of the 3D human body model includes the following steps:

In step S1, a three-dimensional human body model to be edited is obtained.

Further, step S1 also acquires the editing area selected by the user, wherein the embodiment of the present invention can modify the three-dimensional geometry of the selected editing area according to the sketch input by the user.

In step S2, the three-dimensional human body model to be edited is converted into a geometric image by using a surface parameterization method.

Understandably, current 3D neural networks are less mature than 2D neural networks. The three-dimensional information is converted into easy-to-handle two-dimensional information by means of surface parameterization, which is convenient for neural network processing. The reversible conversion method also prevents the information from being lost too much before and after the conversion.

Further, in one embodiment of the present invention, before obtaining the geometric image, one or more vertex chains to be deformed need to be selected in the editing area selected by the user, and a line image is generated by using the vertex chains.

It should be noted that traditional 3D editing methods often require multiple selections of different sketch surfaces for editing at different viewing angles. The embodiment of the present invention only needs to select a vertex chain (composed of multiple vertices connected in a certain order) Automatically generate line images (that is, sketches), no need to define sketch faces, get rid of the limitation of viewing angle, and simplify the editing steps.

In step S3, process the geometric image to train the two-dimensional neural network, output the generated geometric image, and convert the generated geometric image into three-dimensional information.

Specifically, the line image of the geometric figure is extracted through the edge detection algorithm, and a mask image is randomly generated; the mask calculation is performed on the geometric image and the mask image to obtain a two-dimensional incomplete image; train. Input the 3D human body model to be edited into the trained 2D neural network, output the generated geometric image, and convert the 2D information into 3D information. Wherein, the mask image is randomly mapped according to the editing area selected by the user.

It should be noted that the embodiment of the present invention only encodes the difference between the three-dimensional coordinates of the human body model (including clothing details) relative to the human body model (naked). The numerical range of the human body model on the xyz three axes is too different. By only encoding the difference, the numerical range can be unified, and more detailed information is retained, which is convenient for neural network training.

The heuristic editing method of the 3D human body model of the present invention will be further described below with reference to specific examples.

This specific example converts three-dimensional information into a 512×512×3 two-dimensional image. The three image channels are used to store information in the xyz dimension respectively. Since it involves the training and use of the neural network, the following steps are divided into training steps and usage steps.

Training step 1: Use the surface parameterization method to convert the three-dimensional coordinate difference between the human body model (including clothing details) and the human body model (naked) in the data set into a 512×512×3 two-dimensional image (hereinafter referred to as “geometry image").

Training step 2: using edge detection algorithm to extract 512×512×1 line image of geometric image;

Training step 3: Randomly generate a mask image of 512×512×1;

Training step 4: train a two-dimensional neural network, the input of which is a 512×512×3 two-dimensional incomplete image (obtained by the mask calculation of the geometric image and the mask image) and a 512×512×1 line image, and its output Generate geometric images for 512x512x3. During training, in addition to the L1 loss constraint between the original geometric image and the generated geometric image, an adversarial training technique is also used to set up a discriminant network, using the original geometric image as a positive sample and the generated geometric image as a negative sample to improve the accuracy of the generated results. diversity.

Step 1: Select the 3D human body area to be edited;

Step 2: In the above editing area, select one or more vertex chains (consisting of multiple vertices connected in a certain order) that you want to deform;

Using step 3: Using the previously defined surface parameterized mapping, the embodiment of the present invention will automatically map the human body model to be edited into a 512×512×3 geometric image, and automatically map the editing area using step 1 into a 512×512× 1 mask image, automatically map the vertex chain used in step 2 into a 512×512×1 line image, and input the above information to the trained neural network, and output a geometric image;

Step 4: the embodiment of the present invention will automatically convert the geometric image generated by the network into three-dimensional information, so as to obtain the generated three-dimensional human body model.

According to the heuristic editing method of the three-dimensional human body model proposed by the embodiment of the present invention, the three-dimensional information input by the user is converted into easy-to-handle two-dimensional information through the surface parameterization method, and after being processed by the two-dimensional deep neural network, it is reversely transformed into The 3D information is presented, which simplifies the cumbersome 3D editing operations. The graph has nothing to do with the viewing angle, and there is no limit on the number of lines, which can generate the details of the folds of the clothes.

Next, a heuristic editing device for a three-dimensional human body model according to an embodiment of the present invention is described with reference to the accompanying drawings.

Fig. 2 is a schematic structural diagram of a heuristic editing device for a 3D human body model according to an embodiment of the present invention.

As shown in FIG. 2 , the device 10 includes: an acquisition module 100 , a transformation module 200 and a training module 300 .

Wherein, the obtaining module 100 is used for obtaining the 3D human body model to be edited. The conversion module 200 is used to convert the three-dimensional human body model to be edited into a geometric image by using a surface parameterization method. The training module 300 is used to process the geometric image to train the two-dimensional neural network, output the generated geometric image, and invert the generated geometric image into three-dimensional information.

Further, in one embodiment of the present invention, the training module 300 includes: an extraction unit, which is used to extract the line image of the geometric figure through an edge detection algorithm, and randomly generates a mask image; The code image is used for mask calculation to obtain a two-dimensional incomplete image; the training unit is used to train the two-dimensional neural network by using the two-dimensional incomplete image.

Further, in an embodiment of the present invention, the obtaining module is also used to obtain the editing area selected by the user.

Further, in one embodiment of the present invention, before obtaining the geometric image, it is necessary to select one or more vertex chains to be deformed in the editing area selected by the user, and use the vertex chains to generate line images (that is, edit sketches) .

Further, in an embodiment of the present invention, the mask image is mapped out according to the editing area selected by the user.

According to the heuristic editing device of the 3D human body model proposed in the embodiment of the present invention, the 3D information input by the user is converted into easy-to-handle 2D information through the surface parameterization method, and after being processed by the 2D deep neural network, it is reversely transformed into The 3D information is presented, which simplifies the cumbersome 3D editing operations. The graph has nothing to do with the viewing angle, and there is no limit on the number of lines, which can generate the details of the folds of the clothes.

In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be interpreted as indicating or implying relative importance or implicitly specifying the quantity of indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include at least one of these features. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

In the description of this specification, descriptions referring to the terms "one embodiment", "some embodiments", "example", "specific examples", or "some examples" mean that specific features described in connection with the embodiment or example , structure, material or characteristic is included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the described specific features, structures, materials or characteristics may be combined in any suitable manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples and features of different embodiments or examples described in this specification without conflicting with each other.

Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention, those skilled in the art can make the above-mentioned The embodiments are subject to changes, modifications, substitutions and variations.

Claims (8)

1. A heuristic editing method for a three-dimensional human body model, comprising the following steps: Step S1, obtaining the 3D human body model to be edited; Step S2, using a surface parameterization method to convert the 3D human body model to be edited into a geometric image; and Step S3, processing the geometric image to train the two-dimensional neural network, outputting the generated geometric image, and inverting the generated geometric image into three-dimensional information, Described step S3 comprises: Extract the line image of the geometric image by an edge detection algorithm, and randomly generate a mask image; performing mask calculation on the geometric image and the mask image to obtain a two-dimensional incomplete image; The two-dimensional neural network is trained by using the two-dimensional incomplete image. 2. The heuristic editing method for a three-dimensional human body model according to claim 1, wherein the step S1 also acquires the editing area selected by the user. 3. The heuristic editing method of the three-dimensional human body model according to claim 1, wherein, before obtaining the geometric image, one or more vertex chains that want to be deformed need to be selected in the editing area selected by the user , using the vertex chain to generate a line image. 4. The heuristic editing method for a three-dimensional human body model according to claim 2, wherein a mask image is mapped out according to the editing area selected by the user. 5. A heuristic editing device for a three-dimensional human body model, characterized in that it comprises: An acquisition module, configured to acquire a three-dimensional human body model to be edited; A conversion module, for converting the three-dimensional human body model to be edited into a geometric image by using a surface parameterization method; and The training module is used to process the geometric image to train the two-dimensional neural network, output the generated geometric image, and invert the generated geometric image into three-dimensional information, Wherein, the training module includes: An extracting unit is used to extract the line image of the geometric image through an edge detection algorithm, and randomly generate a mask image; A calculation module, configured to perform mask calculation on the geometric image and the mask image to obtain a two-dimensional incomplete image; A training unit, configured to use the two-dimensional incomplete image to train the two-dimensional neural network. 6 . The device for heuristic editing of a three-dimensional human body model according to claim 5 , wherein the acquiring module is further configured to acquire an editing area selected by a user. 7 . 7. The heuristic editing device for a three-dimensional human body model according to claim 5, wherein, before obtaining the geometric image, one or more vertex chains that want to be deformed need to be selected in the editing area selected by the user , using the vertex chain to generate a line image. 8 . The heuristic editing device for a three-dimensional human body model according to claim 6 , wherein a mask image is mapped out according to the editing area selected by the user.

Images