CN110555899A - multi-precision grid refinement method based on CNN cloth wrinkle recognition - Google Patents
multi-precision grid refinement method based on CNN cloth wrinkle recognition Download PDFInfo
- Publication number
- CN110555899A CN110555899A CN201910768724.3A CN201910768724A CN110555899A CN 110555899 A CN110555899 A CN 110555899A CN 201910768724 A CN201910768724 A CN 201910768724A CN 110555899 A CN110555899 A CN 110555899A
- Authority
- CN
- China
- Prior art keywords
- cloth
- grid
- cnn
- precision
- wrinkle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T13/00—Animation
- G06T13/20—3D [Three Dimensional] animation
- G06T13/40—3D [Three Dimensional] animation of characters, e.g. humans, animals or virtual beings
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of computer animation and discloses a multi-precision grid refinement method based on CNN cloth wrinkle recognition, which comprises the steps of firstly establishing a human body model, carrying out animation simulation, extracting a key animation frame and segmenting the wrinkle part of the key animation frame, using the segmented model as the input of a convolutional neural network, and obtaining wrinkle recognition through CNN training; refining the identified wrinkle part by adopting a pixel grid; and finally, converting the refined quadrilateral meshes into triangular meshes, and performing cloth simulation. Compared with the traditional fold identification method based on the calculated curvature, the fold identification method based on the curvature calculation has the advantages that the fold identification speed is increased while the accurate identification of the fold is ensured; the grid is refined by adopting a pixel refinement method, so that the grid is more refined, and the simulated cloth is more vivid.
Description
Technical Field
The invention belongs to the technical field of computer animation, and particularly relates to a multi-precision grid refinement method based on CNN cloth wrinkle recognition.
Background
In recent years, virtual simulation technology has been more and more popular, and the application range of virtual simulation technology is wide, and VR technology appearing in recent years belongs to the category of virtual simulation. The game industry is also a popular field, especially the garment animation effect, and the fidelity of the game directly influences the whole animation effect. Because the cloth is a flexible material, the cloth is easy to deform, and certain wrinkles are generated just because of the bending deformation of the cloth, and the fine and smooth feeling of the animation is influenced by the wrinkles. In order to quickly identify the wrinkle part, some researches adopt a traditional method to identify the bending deformation degree of the cloth by calculating the curvature. In order to make the animation effect more vivid, the cloth is often required to be refined, and the adopted method is multi-precision grid reconstruction.
Machine learning has enabled some complex problems to be solved in a compact way in recent years. Tompson et al propose a DNN-based reasoning process instead of a complex computational process to generate simulation results. Lee et al points out the deficiencies of Tompson et al and propose a rapid and reliable layered cloth simulation method, which combines the traditional physical simulation method with a deep neural network to obtain a finer cloth level; villard et al calculated the angle between the two particles using the spring particle model and refined the mesh by determining the degree of bending. Han et al useAnd a subdivision method, namely refining the cloth.
the method does not use complex curvature calculation any more, but still takes some time to identify the folds, the simulation effect is not better than the simulation effect of the original physical method, and the overall effect is not improved much.
Disclosure of Invention
Aiming at the problems, the invention provides a multi-precision grid refinement method based on CNN cloth wrinkle identification, which is characterized in that the method of a convolutional neural network CNN is used for identifying wrinkles of cloth and then carrying out refinement, so that the aims of rapidly identifying the cloth wrinkles and improving the simulation effect are fulfilled.
in order to achieve the purpose, the invention adopts the following technical scheme.
the multi-precision grid refinement method based on CNN cloth wrinkle identification comprises the following steps:
Step 1, establishing a human body model, carrying out animation simulation, extracting a key animation frame, segmenting a wrinkle part of the key animation frame, and storing the segmented model in a picture format; the key animation frames are distinguished according to the amount of folds, and the key animation frames with more folds are the key animation frames;
step 2, the segmented model is used as the input of a convolutional neural network, and a final recognition result, namely the recognition of the folds, is obtained through the training of the convolutional neural network CNN;
Step 3, adopting pixel grid to refine the identified wrinkle part;
And 4, converting the refined quadrilateral meshes into triangular meshes, and performing cloth simulation.
further, the step 1 of establishing a human body model, and the specific process and following principle of performing animation simulation are as follows:
Step 1.1, calculating the motion state of the cloth motion: let the initial motion state be (x)0,v0)
A·Δv＝b (2)
where x is a 3n × 1 dimensional state vector, v is a 3n × 1 dimensional velocity vector, M is a 3n × 3n mass matrix, f is a force vector, n is the number of particles of the fabric, Δ v is the velocity change, and a and b are:
wherein f is0is an initial force vector, v0The initial speed, Δ t is the time variation, and the target state (x) under no constraint is obtained by the equations (1) - (4)1,v1)；
step 1.2 in the animation simulation process of the human body model, the collision treatment of the cloth and the human body uses global constraint, when the cloth and the human body are about to collide, a constraint set g is added at a mass point which is about to collide,
g(p,p1,p2,p3)＝[(p3-p2)(p1-p2)](p-p2)-d≥0 (5)
Wherein p is a mass point, p1,p2,p3The vertex of the triangular patch of the cloth, d is a controllable correction value;
Step 1.3 when x1When the current constraint set is not met, Δ x is calculated according to equation (9),
The calculation formula (9) of the delta x under the dynamic constraint is obtained by the implicit constraint equations of the formulas (6) to (8),
x(t+Δt)＝x(t)+Δtv(t+Δt) (7)
g(x(t+Δt))＝0 (8)
In a constrained state:
Substituting the formula (9) into the formula (10) to obtain
Equation (11) is the constraint set g in the new state,
Where λ is the Lagrangian multiplier, Δ x is the state vector increment,
Step 1.4 obtaining Δ x from step 1.3, updating x1＝x1+ Δ x, modify constraint set g, calculate
obtain a new state (x)1,v1) Collision and animation simulation between the cloth and the human body are performed according to equations (1) to (12).
furthermore, in the step 2, the segmented model is used as an input of a convolutional neural network, and a final recognition result, namely the recognition of the fold, is obtained through training of the convolutional neural network CNN; the specific training process is as follows:
step 2.1, setting the parameters of the model picture divided in the step 1 to be 227 multiplied by 3 as the input of the convolutional neural network CNN;
step 2.2, setting the network framework of the convolutional neural network CNN into 5 convolutional layers, 3 pooling layers and 3 full-connection layers;
and 2.3, carrying out error analysis to minimize the error of the true value and the predicted value, wherein the formula is as follows:
Where n represents the training case, k represents the dimension, tkIs the true value, ykis a predicted value;
when error Enthe precision of the cloth is up to 98% or more, and the training is finished to obtain output, namely the wrinkle identification of the cloth.
further, the step 3 refines the identified wrinkle part by using a pixel grid; the method specifically comprises the following steps:
firstly, setting a curvature threshold value of grid refinement, then comparing the precision of a pixel grid needing refinement with the set curvature threshold value, and when the precision of the pixel grid is greater than the set curvature threshold value, performing refinement on the pixel grid until the precision reaches the set threshold value; and when the precision of the pixel grid is less than or equal to the set curvature threshold, not performing refinement.
Further, the refinement of the pixel grid is performed in two cases,
(1) When the grid needing to be refined is in the internal area, a quadrilateral grid refinement method is adopted;
(2) when the grid needing to be refined is in the boundary area, fine pixel filling is adopted in the boundary area.
Furthermore, the step 4 converts the refined quadrilateral mesh into a triangular mesh for simulation; specifically, there are two methods:
(1) The quadrilateral mesh is segmented along any diagonal line to obtain a continuous triangular mesh;
(2) Restoring the RGB values of the pixels into three vertexes of a triangle to obtain a continuous triangular mesh;
any one of the methods is selected to convert the quadrilateral mesh into a triangular mesh, and the material distribution simulation is carried out.
Furthermore, the RGB values of the pixels are restored to three vertexes of a triangle, and a triangular mesh is obtained; specifically, the RGB values of the pixels are restored to the coordinate plane as three vertexes of a triangle, the three vertexes are connected to obtain a triangular mesh, and in order to enable the triangular meshes restored by the RGB values of all the pixels to be continuous into a whole, the RGB values of the pixels or the RGB values of the pixels related to the adjacent and discontinuous triangular meshes are approximated.
the invention has the beneficial effects that:
the CNN-based cloth wrinkle identification method is adopted when identifying cloth wrinkles, the CNN-based cloth wrinkle identification method relates to the characteristic extraction of CNN, and the CNN-based cloth wrinkles can well extract characteristics, so that the identification speed is greatly improved in the aspect of identifying wrinkles; secondly, by identifying the wrinkles, the grid is refined by adopting a pixel refinement method, so that the grid is more refined, and the cloth is more vivid. The training speed is integrally improved, and the cloth simulation effect is vivid.
drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is an overall algorithm flow diagram of an embodiment of the invention
FIG. 2 is a schematic diagram of wrinkle identification according to an embodiment of the present invention
FIG. 3 is a schematic diagram of grid refinement according to an embodiment of the present invention
Detailed Description
the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention
examples all other examples, which can be obtained by a person skilled in the art without inventive step, are within the scope of protection of the present invention.
as shown in fig. 1, the multi-precision mesh refinement method based on CNN fabric wrinkle identification includes the following steps:
step 1, establishing a human body model, carrying out animation simulation, extracting a key animation frame, segmenting a wrinkle part of the key animation frame, and storing the segmented model in a picture format; the specific process and the following principle are as follows:
Step 1.1, calculating the motion state of the cloth motion: let the initial motion state be (x)0,v0)
A·Δv＝b (2)
where x is a 3n × 1 dimensional state vector, v is a 3n × 1 dimensional velocity vector, M is a 3n × 3n mass matrix, f is a force vector, n is the number of particles of the fabric, Δ v is the velocity change, and a and b are:
Wherein f is0is an initial force vector, v0the initial speed, delta t is the time variation, and the target state (x) under no constraint is obtained by the formulas (1) - (4)1,v1)；
Step 1.2 in the animation simulation process of the human body model, the collision treatment of the cloth and the human body uses global constraint, when the cloth and the human body are about to collide, a constraint set g is added at a mass point which is about to collide,
g(p,p1,p2,p3)＝[(p3-p2)(p1-p2)](p-p2)-d≥0 (5)
The cloth is composed of a plurality of triangular surface patches, p is a mass point, and p is1,p2,p3The vertex of the cloth triangular patch at the mass point p, and d is a controllable correction value;
step 1.3 when x1When the current constraint set is not met, Δ x is calculated according to equation (9),
The calculation formula (9) of the delta x under the dynamic constraint is obtained by the implicit constraint equations of the formulas (6) to (8),
x(t+Δt)＝x(t)+Δtv(t+Δt) (7)
g(x(t+Δt))＝0 (8)
in a constrained state:
substituting the formula (9) into the formula (10) to obtain
equation (11) is the constraint set g in the new state,
Where λ is the Lagrangian multiplier, Δ x is the state vector increment,
step 1.4 obtaining Δ x from step 1.3, updating x1＝x1+ Δ x, modify constraint set g, calculate
Obtain a new state (x)1,v1) Collision and animation simulation between the cloth and the human body are performed according to equations (1) to (12).
Step 2, the segmented model is used as the input of a convolutional neural network, and a final recognition result, namely the recognition of the folds, is obtained through the training of the convolutional neural network CNN; the specific training process is as follows:
step 2.1, setting the parameters of the model picture divided in the step 1 to be 227 multiplied by 3 as the input of the convolutional neural network CNN; wherein 227 × 227 is the size of the picture, and 3 is a color picture;
Step 2.2, setting the network framework of the convolutional neural network CNN into 5 convolutional layers, 3 pooling layers and 3 full-connection layers;
And 2.3, carrying out error analysis to minimize the error of the true value and the predicted value, wherein the formula is as follows:
where n represents the training case, k represents the dimension, tkIs the true value, ykIs a predicted value;
When error EnThe precision of the cloth is up to 98% or more, and the training is finished to obtain output, namely the wrinkle identification of the cloth. Error E of the present embodimentnthe accuracy of the method reaches 100%, and the wrinkle identification result is shown in figure 2.
step 3, refining the identified wrinkle part by adopting a pixel grid; the method specifically comprises the following steps:
Firstly, setting a curvature threshold value of grid refinement, then comparing the precision of a pixel grid needing refinement with the set curvature threshold value, and when the precision of the pixel grid is greater than the set curvature threshold value, performing refinement on the pixel grid until the precision reaches the set threshold value; and when the precision of the pixel grid is less than or equal to the set curvature threshold, not performing refinement.
the refinement of the pixel grid is performed in two cases,
(1) When the grid needing to be refined is in the internal area, a quadrilateral grid refinement method is adopted;
(2) When the grid needing to be refined is in the boundary area, fine pixel filling is adopted in the boundary area.
Step 4, converting the refined quadrilateral mesh into a triangular mesh for simulation; specifically, there are two methods:
(1) The quadrilateral grids are segmented along any diagonal line to obtain three continuous angular grids;
(2) restoring the RGB values of the pixels into three vertexes of a triangle to obtain a continuous triangular mesh; specifically, the RGB values of the pixels are restored to the coordinate plane as three vertexes of a triangle, the three vertexes are connected to obtain a triangular mesh, and in order to enable the triangular meshes restored by the RGB values of all the pixels to be continuous into a whole, the RGB values of the pixels or the RGB values of the pixels related to the adjacent and discontinuous triangular meshes are approximated.
In this embodiment, a first method is first adopted to segment a quadrilateral mesh along any diagonal to obtain a triangular mesh; the results obtained are shown in FIG. 3.
Meanwhile, the second method is adopted in the embodiment, the RGB values of the pixels are restored into three vertexes of a triangle, and a triangular mesh is obtained; the results obtained are in accordance with FIG. 3.
The embodiment of the invention provides a multi-precision grid refinement method based on CNN cloth wrinkle identification, which comprises the following functional characteristics: increasing constraint by using a collision detection method, updating a motion position by adopting implicit dynamics, and simulating human motion; identifying wrinkles using a classical network of CNNs; the wrinkled portion is refined.
the above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.
Claims (7)
1. The multi-precision grid refinement method based on CNN cloth wrinkle identification is characterized by comprising the following steps:
Step 1, establishing a human body model, carrying out animation simulation, extracting a key animation frame, segmenting a wrinkle part of the key animation frame, and storing the segmented model in a picture format;
Step 2, the segmented model is used as the input of a convolutional neural network, and a final recognition result, namely the recognition of the folds, is obtained through the training of the convolutional neural network CNN;
Step 3, adopting pixel grid to refine the identified wrinkle part;
and 4, converting the refined quadrilateral meshes into triangular meshes, and performing cloth simulation.
2. The multi-precision grid refinement method based on CNN cloth wrinkle identification as claimed in claim 1, wherein the human body model is established in step 1, and the specific process and following principle for animation simulation are as follows:
Step 1.1, calculating the motion state of the cloth motion: let the initial motion state be (x)0,v0)
A·Δv＝b (2)
Where x is a 3n × 1 dimensional state vector, v is a 3n × 1 dimensional velocity vector, M is a 3n × 3n mass matrix, f is a force vector, n is the number of particles of the fabric, Δ v is the velocity change, and a and b are:
wherein f is0Is an initial force vector，v0the initial speed, Δ t is the time variation, and the target state (x) under no constraint is obtained by the equations (1) - (4)1,v1)；
Step 1.2 in the animation simulation process of the human body model, the collision treatment of the cloth and the human body uses global constraint, when the cloth and the human body are about to collide, a constraint set g is added at a mass point which is about to collide,
g(p,p1,p2,p3)＝[(p3-p2)(p1-p2)](p-p2)-d≥0 (5)
wherein p is a mass point, p1,p2,p3the vertex of the triangular patch of the cloth, d is a controllable correction value;
Step 1.3 when x1When the current constraint set is not met, Δ x is calculated according to equation (9),
The calculation formula (9) of the delta x under the dynamic constraint is obtained by the implicit constraint equations of the formulas (6) to (8),
x(t+Δt)＝x(t)+Δtv(t+Δt) (7)
g(x(t+Δt))＝0 (8)
In a constrained state:
substituting the formula (9) into the formula (10) to obtain
Equation (11) is the constraint set g in the new state,
Where λ is the Lagrangian multiplier, Δ x is the state vector increment,
step 1.4 obtaining Δ x from step 1.3, updating x1＝x1+ Δ x, modify constraint set g, calculate
obtain a new state (x)1,v1) Collision and animation simulation between the cloth and the human body are performed according to equations (1) to (12).
3. The multi-precision grid refinement method based on CNN cloth wrinkle identification according to claim 1, characterized in that, in step 2, the segmented model is used as the input of a convolutional neural network, and the final identification result is obtained by training of the convolutional neural network CNN, i.e. the wrinkle identification; the specific training process is as follows:
step 2.1, setting the parameters of the model picture divided in the step 1 to be 227 multiplied by 3 as the input of the convolutional neural network CNN;
step 2.2, setting the network framework of the convolutional neural network CNN into 5 convolutional layers, 3 pooling layers and 3 full-connection layers;
And 2.3, carrying out error analysis to minimize the error of the true value and the predicted value, wherein the formula is as follows:
where n represents the training case, k represents the dimension, tkIs the true value, ykis a predicted value;
when error Enthe precision of the cloth is up to 98% or more, and the training is finished to obtain output, namely the wrinkle identification of the cloth.
4. The CNN cloth wrinkle identification-based multi-precision grid refinement method according to claim 1, characterized in that said step 3 refines the identified wrinkle part by using pixel grid; the method specifically comprises the following steps:
firstly, setting a curvature threshold value of grid refinement, then comparing the precision of a pixel grid needing refinement with the set curvature threshold value, and when the precision of the pixel grid is greater than the set curvature threshold value, performing refinement on the pixel grid until the precision reaches the set threshold value; and when the precision of the pixel grid is less than or equal to the set curvature threshold, not performing refinement.
5. The CNN cloth wrinkle identification-based multi-precision grid refinement method according to claim 4, wherein the pixel grid is refined in two cases,
(1) when the grid needing to be refined is in the internal area, a quadrilateral grid refinement method is adopted;
(2) When the grid needing to be refined is in the boundary area, fine pixel filling is adopted in the boundary area.
6. the multi-precision mesh refinement method based on CNN cloth wrinkle identification as claimed in claim 1, wherein said step 4 converts the refined quadrilateral mesh into triangular mesh for simulation; specifically, there are two methods:
(1) The quadrilateral mesh is segmented along any diagonal line to obtain a continuous triangular mesh;
(2) Restoring the RGB values of the pixels into three vertexes of a triangle to obtain a continuous triangular mesh;
any one of the methods is selected to convert the quadrilateral mesh into a triangular mesh, and the material distribution simulation is carried out.
7. The multi-precision mesh refinement method based on CNN cloth wrinkle identification as claimed in claim 6, wherein the RGB values of the pixels are restored to three vertices of a triangle to obtain a triangular mesh; specifically, the RGB values of the pixels are restored to the coordinate plane as three vertexes of a triangle, the three vertexes are connected to obtain a triangular mesh, and in order to enable the triangular meshes restored by the RGB values of all the pixels to be continuous into a whole, the RGB values of the pixels or the RGB values of the pixels related to the adjacent and discontinuous triangular meshes are approximated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910768724.3A CN110555899B (en) | 2019-08-20 | 2019-08-20 | Multi-precision grid refinement method based on CNN cloth wrinkle recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910768724.3A CN110555899B (en) | 2019-08-20 | 2019-08-20 | Multi-precision grid refinement method based on CNN cloth wrinkle recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110555899A true CN110555899A (en) | 2019-12-10 |
| CN110555899B CN110555899B (en) | 2022-09-16 |
Family
ID=68737710
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910768724.3A Active CN110555899B (en) | 2019-08-20 | 2019-08-20 | Multi-precision grid refinement method based on CNN cloth wrinkle recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110555899B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112767553A (en) * | 2021-02-02 | 2021-05-07 | 华北电力大学 | Self-adaptive group clothing animation modeling method |
| CN114170354A (en) * | 2021-11-03 | 2022-03-11 | 完美世界(北京)软件科技发展有限公司 | Virtual character clothing manufacturing method, device, equipment, program and readable medium |
| CN114662172A (en) * | 2022-05-19 | 2022-06-24 | 武汉纺织大学 | Garment fabric dynamic simulation method based on neural network |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018011649A1 (en) * | 2016-07-14 | 2018-01-18 | Moda-Match Ltd. | Fitting clothing articles to human images |
| US20190043269A1 (en) * | 2017-08-03 | 2019-02-07 | The University Of North Carolina At Chapel Hill | Methods, systems, and computer readable media for modeling garments using single view images |
| CN109509171A (en) * | 2018-09-20 | 2019-03-22 | 江苏理工学院 | A kind of Fabric Defects Inspection detection method based on GMM and image pyramid |
| CN109934287A (en) * | 2019-03-12 | 2019-06-25 | 上海宝尊电子商务有限公司 | A kind of clothing texture method for identifying and classifying based on LBP and GLCM |
| CN109961022A (en) * | 2019-03-07 | 2019-07-02 | 首都师范大学 | A kind of clothes sketch input fabric material identification analogy method based on deep learning |
-
2019
- 2019-08-20 CN CN201910768724.3A patent/CN110555899B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018011649A1 (en) * | 2016-07-14 | 2018-01-18 | Moda-Match Ltd. | Fitting clothing articles to human images |
| US20190043269A1 (en) * | 2017-08-03 | 2019-02-07 | The University Of North Carolina At Chapel Hill | Methods, systems, and computer readable media for modeling garments using single view images |
| CN109509171A (en) * | 2018-09-20 | 2019-03-22 | 江苏理工学院 | A kind of Fabric Defects Inspection detection method based on GMM and image pyramid |
| CN109961022A (en) * | 2019-03-07 | 2019-07-02 | 首都师范大学 | A kind of clothes sketch input fabric material identification analogy method based on deep learning |
| CN109934287A (en) * | 2019-03-12 | 2019-06-25 | 上海宝尊电子商务有限公司 | A kind of clothing texture method for identifying and classifying based on LBP and GLCM |
Non-Patent Citations (2)
| Title |
|---|
| LAN CHEN等: ""Synthesizing Cloth Wrinkles by CNN-based Geometry Image Super-resolution"", 《COMPUTER ANIMATION AND VIRTUAL WORLDS》 * |
| 石敏等: ""基于实例数据分析的多精度网格布料动画"", 《计算机学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112767553A (en) * | 2021-02-02 | 2021-05-07 | 华北电力大学 | Self-adaptive group clothing animation modeling method |
| CN114170354A (en) * | 2021-11-03 | 2022-03-11 | 完美世界(北京)软件科技发展有限公司 | Virtual character clothing manufacturing method, device, equipment, program and readable medium |
| CN114170354B (en) * | 2021-11-03 | 2022-08-26 | 完美世界(北京)软件科技发展有限公司 | Virtual character clothing manufacturing method, device, equipment, program and readable medium |
| CN114662172A (en) * | 2022-05-19 | 2022-06-24 | 武汉纺织大学 | Garment fabric dynamic simulation method based on neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110555899B (en) | 2022-09-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110555899B (en) | Multi-precision grid refinement method based on CNN cloth wrinkle recognition | |
| Jiang et al. | Bcnet: Learning body and cloth shape from a single image | |
| CN104008564B (en) | A kind of human face expression cloning process | |
| CN106570867A (en) | ACM (Active Contour Model) image rapid segmentation method based on gray scale morphological energy method | |
| CN109064522A (en) | The Chinese character style generation method of confrontation network is generated based on condition | |
| CN106530317B (en) | A kind of scoring of simple picture computer and auxiliary painting methods | |
| CN104778736B (en) | The clothes three-dimensional animation generation method of single video content driven | |
| CN109002630B (en) | Rapid simulation method for super-elastic material | |
| CN104809457A (en) | Three-dimensional face identification method and system based on regionalization implicit function features | |
| CN103914862A (en) | Pencil sketch simulating method based on edge tangent stream | |
| CN114662172A (en) | Garment fabric dynamic simulation method based on neural network | |
| CN108460781A (en) | A kind of active contour image partition method and device based on improvement SPF | |
| CN114820294A (en) | All-dimensional virtual fitting method, system and medium based on cyclic three-level transformation | |
| Koh et al. | View-dependent adaptive cloth simulation | |
| Shao et al. | Towards Multi-Layered 3D Garments Animation | |
| CN111047684A (en) | Model simplification method based on three-dimensional model characteristics | |
| US11869163B1 (en) | Virtual garment draping using machine learning | |
| CN104517299A (en) | Method for restoring and resimulating physical video fluid driving model | |
| Achar et al. | A Comparative Study of Garment Draping Techniques | |
| Liu et al. | Robust and automatic clothing reconstruction based on a single RGB image | |
| CN104156502B (en) | A kind of location-based clothing fold geometry generation method | |
| Li et al. | Garment Recovery with Shape and Deformation Priors | |
| Li et al. | High-quality animatable dynamic garment reconstruction from monocular videos | |
| Li et al. | 3D clothing fitting based on the geometric feature matching | |
| CN116580164B (en) | Dressing feature learning method for single-view three-dimensional human body reconstruction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |