CN115795627A - Furniture feature construction method, system, device and medium - Google Patents
Furniture feature construction method, system, device and medium Download PDFInfo
- Publication number
- CN115795627A CN115795627A CN202211693388.9A CN202211693388A CN115795627A CN 115795627 A CN115795627 A CN 115795627A CN 202211693388 A CN202211693388 A CN 202211693388A CN 115795627 A CN115795627 A CN 115795627A
- Authority
- CN
- China
- Prior art keywords
- product
- information
- furniture
- feature
- characteristic
- 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
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a furniture feature construction method, a system, a device and a medium, wherein the method comprises the following steps: acquiring first product information of alternative products in a product library, and constructing a product characteristic matrix and a process characteristic matrix according to the first product information; acquiring room characteristic information in a historical design scheme, and training the furniture characteristic pre-training model by using a product characteristic matrix, a process characteristic matrix and the room characteristic information; the method comprises the steps of obtaining second product information and second process information of a target product, inputting the second product information and the second process information into a furniture feature pre-training model after training, and outputting and obtaining a room type matched with the target product.
Description
Technical Field
The invention relates to the technical field of feature engineering, in particular to a furniture feature construction method, a system, a device and a medium.
Background
The traditional machine learning function development needs a complex characteristic engineering process, namely, a developer combines artificial priori knowledge to search, calculate and establish various characteristics of the function in the field range, and the development of specific functions is completed by utilizing the characteristics through a machine learning model. In addition, in the actual application process, the constructed feature engineering usually needs to be verified by repeatedly testing a large number of features, and the process is relatively tedious and inefficient.
Disclosure of Invention
In view of the above, in order to at least partially solve one of the above technical problems or disadvantages, an embodiment of the present invention provides a furniture feature construction method, which aims to reduce the difficulty in developing feature functions, construct a feature library through a pre-training model, and do not need to perform too many complex feature projects in the subsequent machine learning function development; the technical scheme of the application also provides a system, a device and a medium corresponding to the method.
On one hand, the technical scheme of the application provides a furniture feature construction method, which comprises the following steps:
acquiring first product information of alternative products in a product library, and constructing a product characteristic matrix according to the first product information;
traversing the first product information, extracting to obtain first process information corresponding to the alternative product, and constructing a process characteristic matrix according to the first process information;
acquiring a historical design scheme, and extracting room characteristic information in the historical design scheme, wherein the historical design scheme comprises at least one alternative product in the product library;
inputting the product characteristic matrix, the process characteristic matrix and the room characteristic information into a furniture characteristic pre-training model, and training the furniture characteristic pre-training model;
and acquiring second product information and second process information of a target product, inputting the second product information and the second process information into a trained furniture feature pre-training model, and outputting to obtain a room type matched with the target product.
In a possible embodiment of the present application, before the step of obtaining first product information of an alternative product in a product library and constructing a product feature matrix according to the first product information, the method includes:
determining product items according to the candidate products with the finest granularity, and constructing the product library according to the product items;
and traversing the product library, and deleting the repeated product items and the wrong product items in the product library.
In a feasible embodiment of the scheme of the application, the step of traversing the first product information, extracting first process information corresponding to an alternative product, and constructing a process feature matrix according to the first process information includes:
obtaining a product process document corresponding to the alternative product in the product library;
dividing the process attributes recorded in the product process document to obtain the discrete attributes and the continuous attributes;
and constructing the process characteristic matrix according to the discrete attributes and the continuous attributes.
In one possible embodiment of the present solution, the continuous attribute includes a discrete portion and an undiscreted portion; before the step of constructing the process feature matrix according to the discrete attributes and the continuous attributes, the method further comprises the following steps:
discretizing the continuous attribute to obtain a discrete attribute;
the discretization processing comprises the following steps:
preserving the discrete portion of the continuous attribute;
performing discrete slicing on the undispersed part according to the discrete part to obtain a discrete interval;
and integrating the discrete interval and the discrete part to obtain the discrete attribute.
In a possible embodiment of the present disclosure, the obtaining the historical design solution and extracting the room feature information in the historical design solution includes at least one of the following steps:
extracting wall characteristic information and wall position information in the room information;
extracting furniture feature information and furniture position information in the room information;
and constructing to obtain the room characteristic information according to the wall characteristic information, the wall position information, the furniture characteristic information and the furniture position information.
In a possible embodiment of the present disclosure, the inputting the product feature matrix, the process feature matrix, and the room feature information into a furniture feature pre-training model, and the training the furniture feature pre-training model includes:
obtaining a wall input vector according to a product characteristic matrix and a process characteristic matrix combination corresponding to the wall in the room characteristic information;
combining to obtain furniture input vectors according to product characteristic matrixes and process characteristic matrixes corresponding to furniture in the room characteristic information;
and combining the room types, the wall input vectors and the furniture input vectors in the room characteristic information to obtain characteristic combination vectors, and training the furniture characteristic pre-training model according to the characteristic combination vectors.
In a feasible embodiment of the present application, the combining the room type, the wall input vector, and the furniture input vector in the room feature information to obtain a feature combination vector, and training the furniture feature pre-training model according to the model input vector includes:
replacing partial characters in the feature combination vector through a mask character to obtain a model input vector;
obtaining the model input vector, inputting the model input vector to the furniture feature pre-training model and outputting the model input vector to obtain a model output vector;
calculating to obtain cross entropy loss of the quality inspection of the model input vector and the model output vector, and training the furniture feature pre-training model according to the cross entropy loss;
and determining the cross entropy loss convergence to obtain a trained furniture feature pre-training model.
On the other hand, the technical scheme of the application also provides a furniture feature construction system, which comprises:
the product feature extraction unit is used for acquiring first product information of alternative products in a product library and constructing a product feature matrix according to the first product information; the first process information is used for traversing the first product information, extracting first process information corresponding to the alternative product, and constructing a process characteristic matrix according to the first process information;
the house characteristic extraction unit is used for acquiring a historical design scheme and extracting room characteristic information in the historical design scheme, wherein the historical design scheme comprises at least one alternative product in the product library;
the model training unit is used for inputting the product characteristic matrix, the process characteristic matrix and the room characteristic information into a furniture characteristic pre-training model and training the furniture characteristic pre-training model;
and the downstream application unit is used for acquiring second product information and second process information of a target product, inputting the second product information and the second process information into a trained furniture feature pre-training model, and outputting to obtain the room type matched with the target product.
On the other hand, the technical scheme of the application also provides a furniture feature constructing device, which comprises at least one processor; at least one memory for storing at least one program; when executed by the at least one processor, cause the at least one processor to execute a furniture feature construction method as described in the first aspect.
In another aspect, the present technical solution also provides a storage medium, in which a processor-executable program is stored, and the processor-executable program is used for executing the furniture feature construction method according to any one of the first aspect when executed by a processor.
Advantages and benefits of the present 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:
according to the technical scheme, the furniture pre-training model is constructed, the properly matched furniture information is combined together through the historical furniture design scheme, the furniture feature library with the furniture process attribute and the actual matching knowledge is formed, and the process is completely carried out on line. By using the furniture feature library, the feature engineering time of a downstream intelligent home design task can be greatly shortened, the task development efficiency is improved, and meanwhile, the effect of the downstream task can also be improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the description below are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
Fig. 1 is a flowchart illustrating steps of a furniture feature constructing method according to an embodiment of the present disclosure;
fig. 2 is a model schematic diagram of a furniture feature construction method based on a pre-training model in the technical scheme of the application.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention. For the step numbers in the following embodiments, they are set for convenience of illustration only, the order between the steps is not limited at all, and the execution order of each step in the embodiments can be adapted according to the understanding of those skilled in the art.
It should be noted that although functional blocks are partitioned in a schematic diagram of an apparatus and a logical order is shown in a flowchart, in some cases, the steps shown or described may be performed in a different order than the partitioning of blocks in the apparatus or the order in the flowchart. The terms first, second and the like in the description and in the claims, as well as in the drawings described above, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the present application only and is not intended to be limiting of the application.
Before describing the specific embodiments of the present technical solution, it is necessary to make necessary definitions for some characteristic terms:
and the feature library provides a single window for sharing the digital features of the real objects in the specific field. When a new project is started, the required features can be easily found in the feature library.
Based on the technical theoretical basis, in the related technical scheme, particularly in the field of deep learning, the advanced feature library construction technology is adopted; for example, the BERT model may pre-train text features according to left and right bi-directional contexts for text corpora; the MAE model can pre-train image features for a two-dimensional image by a method for monitoring the pixels of the image; the UNITER model combines the information of a plurality of modes of texts and images to perform a pre-training task to obtain the characteristics of the combined multi-mode; the FashionBERT model adopted by E-business software in industrial application is also similar to UNITER in a method for constructing commodity characteristics in E-business scenes.
Aiming at the technical defects and combining the vacancy in the field of home design, the technical scheme of the application integrates the existing pre-training model technology in academic circles into the field of customized furniture and home design, and the constructed furniture features can be well applied to the downstream tasks of home design. The use scenario of the feature library finally constructed by the technical solution of the present application may include, but is not limited to, the following application scenarios: furniture intelligent layout, furniture placement recommendation, intelligent ornament placement and the like.
With reference to fig. 1 in the specification, for a furniture feature construction method provided in an embodiment of the present application, the method may include steps S100 to S500:
s100, first product information of alternative products in a product library is obtained, and a product characteristic matrix is constructed according to the first product information.
In particular embodiments, embodiments may retrieve all product totals from a product database. In some possible embodiments, before step S100 of the method provided by the embodiment, it is necessary to pre-construct a product library containing all products, and before step S100, the embodiment may further include steps S001 to S002:
s001, determining product items according to a plurality of candidate products with the finest granularity, and constructing the product library according to the product items;
and S002, traversing the product library, and deleting repeated product items and wrong product items in the product library.
In particular, in the embodiment, the product material records accumulated by the furniture manufacturing company can be firstly recorded, and then a product finished product with the finest granularity in the product material records is taken as a product item, rather than a product family, an assembly, a component and a part; for example, the single high-voltage electric appliance floor cabinet is a product family and cannot be used as one product item, and the handle-free single high-voltage electric appliance floor cabinet and the bright handle single high-voltage electric appliance floor cabinet can be used as two product items. Further, the examples combine all product items into the full product library of the examples. On the basis of constructing the product full-scale library, the embodiment traverses the product full-scale library and deletes all repeated product items and wrong product items; wherein, the repeated product items can be product items generated in the design and development process of different series of products; the wrong product item may be a product item that violates a process constraint during subsequent production development.
S200, traversing the first product information, extracting to obtain first process information corresponding to the alternative product, and constructing a process characteristic matrix according to the first process information.
Specifically, in the embodiment, the product total quantity library constructed in step 100 is obtained, the product total quantity library is traversed, a product process document corresponding to each product is obtained, and main product process attributes such as a product configurable range (length, width, and height, respectively), a product configurable color, a product configurable part (such as a door panel of a cabinet, a drawer of the cabinet, a handle of a wooden door, an armrest of a chair, and the like) are recorded.
In some optional embodiments, the first process information includes a discrete attribute and a continuous attribute, and step S200 in the embodiment may include steps S210 to S230:
s210, obtaining a product process document corresponding to the alternative product in the product library;
s220, dividing the process attributes recorded in the product process document to obtain the discrete attributes and the continuous attributes;
s230, constructing the process characteristic matrix according to the discrete attributes and the continuous attributes.
Specifically, in the embodiment, for the product process attribute, the discrete attribute value and the continuous attribute value are separately recorded; for example, the product manufacturable height may be recorded as a discrete product manufacturable height, a continuous product manufacturable height; and (3) summarizing a process attribute total library, wherein the final structural form is as follows:
{
discrete product manufacturable lengths: 5, 18, 20, 50, 100, 150.;
the continuous product can be manufactured in length: 100-1000, 50-2800.;
the product can be made into flower color: precious gold, baculou bronze, thousand woven silver wires and enamel ash;
......；
}。
further, in some optional embodiments, before the step S230 of constructing the process feature matrix according to the discrete attributes and the continuous attributes, the embodiment may further perform discretization on the continuous attributes to obtain the discrete attributes; in this embodiment, the process of discretization processing may include steps S231-S233:
s231, reserving the discrete part in the continuous attribute;
s232, performing discrete slicing on the undispersed part according to the dispersed part to obtain a discrete interval;
s233, integrating the discrete interval and the discrete part to obtain the discrete attribute.
In the embodiment, each continuous attribute is divided into a discrete part and an undistributed part, and special discretization processing is required; firstly, the discrete part is reserved and not processed as the discrete part is a common value; then, the embodiment carries out discrete segmentation on the part which is not discrete according to the discrete part, and the interval which is not the largest is set to be not more than 10; and finally integrating the values of the two parts into a complete discrete attribute.
S300, acquiring a historical design scheme, and extracting room characteristic information in the historical design scheme; wherein at least one alternative product in the product library is included in the historical design solution.
In a specific embodiment, the historical design scheme is traversed, and all the wall, door and window and furniture information in the historical design scheme is obtained and used as training data of a subsequent furniture feature pre-training model. The whole house customized design scheme and the single customized cabinet design scheme of the customized home history can be obtained in the embodiment, and the whole house customized design scheme and the single customized cabinet design scheme comprise the cabinet body and the furniture placement information around the cabinet body.
In some optional embodiments, the example method step S300 may include steps S310-S330:
s310, extracting wall characteristic information and wall position information in the room information;
s320, extracting furniture feature information and furniture position information in the room information;
s330, the room characteristic information is constructed and obtained according to the wall characteristic information, the wall position information, the furniture characteristic information and the furniture position information.
In particular, in an embodiment, a list of room types may be constructed and initialized; then, traversing all the design schemes, and respectively carrying out the following operations:
A. recording characteristic information of all walls, such as wall length, wall thickness, wall height, inner surface patterns of the walls, outer surface patterns of the walls and the like;
B. recording position information of all the wall bodies, namely a left horizontal axis coordinate, a rear vertical axis coordinate and a bottom height coordinate of the wall bodies on the three-dimensional space;
C. recording characteristic information of all furniture, such as furniture length, furniture depth, furniture height, furniture pattern, furniture fittings and the like;
D. recording position information of all furniture, namely left horizontal axis coordinates, rear vertical axis coordinates and bottom height coordinates of the furniture in the three-dimensional space;
E. according to the position information of all the walls and the position information of all the furniture, finding out a point with the minimum coordinate value as a starting point of the scheme, and uniformly subtracting the starting point of the scheme from all the position coordinate points;
F. the type of each room is recorded into a room type list.
S400, inputting the product characteristic matrix, the process characteristic matrix and the room characteristic information into a furniture characteristic pre-training model, and training the furniture characteristic pre-training model.
Specifically, in the embodiment, an initialized product feature matrix is constructed in step S100, and is denoted as Eproduct, the feature dimension of each product is set to 256, and if the total number of the product full-scale library is N, the size of the product feature matrix is (N + 3) × 256, where the two extra entries are a separator, a terminator, and a [ MASK ] symbol. Acquiring a process attribute full-scale library in step S200, which is marked as eattributton, constructing an initialized attribute feature matrix for each process attribute, which is marked as eattributton [ a ] for process a and the feature dimension of each attribute is set to 256, if the total discrete number of a certain attribute is K, the size of the attribute feature matrix is K × 256, and the total number of the process attributes is M. And building an initialized room type feature matrix according to the room type list built in the step S300, and recording the initialized room type feature matrix as Eroom, wherein the feature dimension of each room type is set to be 256, and if the total number of the room types is F, the size of the room type feature matrix is F × 256. And combining the EAttribute, the EAttribute [ A ] and the Eroom to obtain an input vector of the furniture feature pre-training model.
Further, in the embodiment, the furniture feature pre-training model may adopt a BERT model, the depth of the model is 12 layers, the feature dimension is 256, and during the model training process, especially during the process of inputting training data into the model, steps S410-S430 may be included:
s410, combining a product characteristic matrix and a process characteristic matrix corresponding to a wall body in the room characteristic information to obtain a wall body input vector;
s420, combining to obtain furniture input vectors according to product feature matrixes and process feature matrixes corresponding to the furniture in the room feature information;
s430, combining the room types, the wall input vectors and the furniture input vectors in the room characteristic information to obtain characteristic combination vectors, and training the furniture characteristic pre-training model according to the characteristic combination vectors.
Specifically, in the embodiment, as shown in fig. 2, in the embodiment, the BERT model input is a room type, all walls, and all furniture placed in a room in a history scheme, where the wall and the furniture are divided into a plurality of inputs according to the number of their attributes, and the room type has only one input (a complete input such as [ wall (length), wall (thickness), wall (height), separator,. ·, furniture 1 (length), furniture 1 (width),. ·, separator, room type, terminator ]), where each entry of one input is a feature vector with a dimension of 256, denoted as Eobj, added by Eproduct, EAttribute [ a ], and Eposition, where Eposition is position coding.
Still further, step S430 in an embodiment method may include steps S431-S434:
s431, replacing partial characters in the feature combination vector through a mask character to obtain a model input vector;
s432, obtaining the model input vector, inputting the model input vector to the furniture feature pre-training model, and outputting the model input vector to obtain a model output vector;
s433, calculating to obtain cross entropy loss of the quality inspection of the model input vector and the model output vector, and training the furniture feature pre-training model according to the cross entropy loss;
and S434, determining the cross entropy loss convergence to obtain a trained pre-training model for implementing furniture features.
In particular, in the embodiment, in addition to the room type, the separator, and the terminator, the input of the remaining 15% is randomly picked and uniformly replaced with the feature vector of the [ MASK ] character. And then, performing loss calculation on the output of the model, wherein only the loss is calculated on the output corresponding to the [ MASK ] symbol in the input, and the loss value is calculated by adopting cross entropy loss. In addition, the model in the embodiment may employ Adam optimizer, initial learning rate of 0.0001, and step-down learning rate decay method, with batch size defined as 64 rooms, for a total of 100 rounds of training.
S500, obtaining second product information and second process information of a target product, inputting the second product information and the second process information into a trained furniture feature pre-training model, and outputting to obtain a room type matched with the target product.
Specifically, in an embodiment, a product feature matrix, a process attribute feature matrix set, a room type feature matrix, and a pre-training model are saved. Storing a product characteristic matrix with the size of N256; saving a process attribute feature matrix set, wherein the number of the process attribute feature matrices is M, and the size of each matrix is K × 256; saving a room type feature matrix with the size of F × 256; the pre-training model is saved.
In the downstream application of the embodiment, a combination of the product feature and the process attribute feature is used as the object feature. Firstly, reading a product feature matrix; then, reading a process attribute feature matrix set; acquiring a product characteristic P from the product characteristic matrix according to the product index; acquiring a process attribute characteristic A from a corresponding process attribute characteristic matrix according to the process attribute required by the downstream machine learning task and the index; and finally, completing a machine learning task according to the product characteristics P and the process attribute characteristics A.
On the other hand, the technical scheme of the application also provides a furniture feature construction system, which comprises:
the product feature extraction unit is used for acquiring first product information of alternative products in a product library and constructing a product feature matrix according to the first product information; the first process information corresponding to the alternative product is extracted and obtained by traversing the first product information, and a process characteristic matrix is constructed according to the first process information;
the house characteristic extraction unit is used for acquiring a historical design scheme and extracting room characteristic information in the historical design scheme, wherein the historical design scheme comprises at least one alternative product in the product library;
the model training unit is used for inputting the product characteristic matrix, the process characteristic matrix and the room characteristic information into a furniture characteristic pre-training model and training the furniture characteristic pre-training model;
and the downstream application unit is used for acquiring second product information and second process information of a target product, inputting the second product information and the second process information into a trained furniture feature pre-training model, and outputting to obtain the room type matched with the target product.
On the other hand, this application technical scheme still provides a furniture characteristic construction device, and the device includes: at least one processor; at least one memory for storing at least one program; when executed by the at least one processor, cause the at least one processor to execute a furniture feature construction method according to the second aspect.
An embodiment of the present invention further provides a storage medium, where a corresponding execution program is stored, and the program is executed by a processor, so as to implement the furniture feature construction method in the first aspect.
From the above specific implementation process, it can be concluded that the technical solution provided by the present invention has the following advantages or advantages compared to the prior art:
according to the technical scheme, the furniture pre-training model is constructed, the properly matched furniture information is combined together through the historical furniture design scheme, the furniture feature library with the furniture process attribute and the actual matching knowledge is formed, and the process is completely carried out on line. By using the furniture feature library, the feature engineering time of a downstream intelligent home design task can be greatly shortened, the task development efficiency is improved, and meanwhile, the effect of the downstream task can also be improved.
Furthermore, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified to the contrary, one or more of the functions and/or features may be integrated in a single physical device and/or software module, or one or more functions and/or features may be implemented in separate physical devices or software modules. It will also be understood that a detailed discussion of the actual implementation of each module is not necessary for an understanding of the present invention. Rather, the actual implementation of the various functional modules in the apparatus disclosed herein will be understood within the ordinary skill of an engineer, given the nature, function, and internal relationship of the modules. Accordingly, those of ordinary skill in the art will be able to practice the invention as set forth in the claims without undue experimentation. It is also to be understood that the specific concepts disclosed are merely illustrative of and not intended to limit the scope of the invention, which is defined by the appended claims and their full scope of equivalents.
The logic and/or steps represented in the flowcharts or otherwise described herein, e.g., an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. A furniture feature construction method is characterized by comprising the following steps:
acquiring first product information of alternative products in a product library, and constructing a product characteristic matrix according to the first product information;
traversing the first product information, extracting to obtain first process information corresponding to the alternative product, and constructing a process characteristic matrix according to the first process information;
acquiring a historical design scheme, and extracting room characteristic information in the historical design scheme, wherein the historical design scheme comprises at least one alternative product in the product library;
inputting the product characteristic matrix, the process characteristic matrix and the room characteristic information into a furniture characteristic pre-training model, and training the furniture characteristic pre-training model;
and acquiring second product information and second process information of a target product, inputting the second product information and the second process information into a trained furniture feature pre-training model, and outputting to obtain the room type matched with the target product.
2. The furniture feature construction method according to claim 1, wherein before the step of obtaining first product information of candidate products in a product library and constructing a product feature matrix according to the first product information, the method comprises:
determining product items according to the candidate products with the finest granularity, and constructing the product library according to the product items;
and traversing the product library, and deleting repeated product items and wrong product items in the product library.
3. The furniture feature construction method according to claim 1, wherein the first process information includes discrete attributes and continuous attributes, the step of traversing the first product information, extracting first process information corresponding to the candidate product, and constructing a process feature matrix according to the first process information includes:
obtaining a product process document corresponding to the alternative product in the product library;
dividing the process attributes recorded in the product process document to obtain the discrete attributes and the continuous attributes;
and constructing the process characteristic matrix according to the discrete attributes and the continuous attributes.
4. A furniture feature construction method according to claim 3, wherein said continuous properties comprise discrete parts and non-discrete parts; before the step of constructing the process feature matrix according to the discrete attributes and the continuous attributes, the method further comprises the following steps:
discretizing the continuous attribute to obtain a discrete attribute;
the discretization processing comprises the following steps:
preserving the discrete part in the continuous attribute;
performing discrete slicing on the undispersed part according to the discrete part to obtain a discrete interval;
and integrating the discrete interval and the discrete part to obtain the discrete attribute.
5. The furniture feature construction method according to claim 1, wherein the obtaining of the historical design solution and the extracting of the room feature information in the historical design solution comprise at least one of the following steps:
extracting wall characteristic information and wall position information in the room information;
extracting furniture feature information and furniture position information in the room information;
and constructing to obtain the room characteristic information according to the wall characteristic information, the wall position information, the furniture characteristic information and the furniture position information.
6. The furniture feature construction method according to claim 5, wherein the inputting the product feature matrix, the process feature matrix and the room feature information into a furniture feature pre-training model, and the training the furniture feature pre-training model comprises:
obtaining a wall input vector according to a product characteristic matrix and a process characteristic matrix combination corresponding to the wall in the room characteristic information;
combining to obtain furniture input vectors according to product characteristic matrixes and process characteristic matrixes corresponding to furniture in the room characteristic information;
and combining the room types, the wall input vectors and the furniture input vectors in the room characteristic information to obtain characteristic combination vectors, and training the furniture characteristic pre-training model according to the characteristic combination vectors.
7. The furniture feature construction method according to claim 6, wherein the combining according to the room type, the wall input vector and the furniture input vector in the room feature information to obtain a feature combination vector, and the training according to the model input vector on the furniture feature pre-training model comprises:
replacing partial characters in the feature combination vector through a mask character to obtain a model input vector;
obtaining the model input vector, inputting the model input vector to the furniture feature pre-training model and outputting the model input vector to obtain a model output vector;
calculating to obtain cross entropy loss of the quality inspection of the model input vector and the model output vector, and training the furniture feature pre-training model according to the cross entropy loss;
and determining the cross entropy loss convergence to obtain a trained furniture feature pre-training model.
8. A furniture feature construction system, comprising:
the product feature extraction unit is used for acquiring first product information of alternative products in a product library and constructing a product feature matrix according to the first product information; the first process information is used for traversing the first product information, extracting first process information corresponding to the alternative product, and constructing a process characteristic matrix according to the first process information;
the house characteristic extraction unit is used for acquiring a historical design scheme and extracting room characteristic information in the historical design scheme, wherein the historical design scheme comprises at least one alternative product in the product library;
the model training unit is used for inputting the product characteristic matrix, the process characteristic matrix and the room characteristic information into a furniture characteristic pre-training model and training the furniture characteristic pre-training model;
and the downstream application unit is used for acquiring second product information and second process information of a target product, inputting the second product information and the second process information into a trained furniture feature pre-training model, and outputting to obtain a room type matched with the target product.
9. A furniture feature construction device, characterized in that the device comprises:
at least one processor;
at least one memory for storing at least one program;
when executed by the at least one processor, cause the at least one processor to perform a furniture feature construction method according to any one of claims 1-7.
10. A storage medium in which a processor-executable program is stored, wherein the processor-executable program, when executed by a processor, is adapted to perform a furniture feature construction method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211693388.9A CN115795627B (en) | 2022-12-28 | 2022-12-28 | Furniture feature construction method, system, device and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211693388.9A CN115795627B (en) | 2022-12-28 | 2022-12-28 | Furniture feature construction method, system, device and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115795627A true CN115795627A (en) | 2023-03-14 |
| CN115795627B CN115795627B (en) | 2023-09-26 |
Family
ID=85427050
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211693388.9A Active CN115795627B (en) | 2022-12-28 | 2022-12-28 | Furniture feature construction method, system, device and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115795627B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200203017A1 (en) * | 2017-06-09 | 2020-06-25 | Alerte Digital Sport Pty Ltd | Systems and methods of prediction of injury risk with a training regime |
| US20210064792A1 (en) * | 2019-08-27 | 2021-03-04 | Lg Electronics Inc. | Artificial intelligence device for guiding furniture placement and method of operating the same |
| CN113538101A (en) * | 2021-07-26 | 2021-10-22 | 杨浩松 | Method, system and storage medium for customizing home decoration based on artificial intelligence |
| CN113743009A (en) * | 2021-08-31 | 2021-12-03 | 广州极点三维信息科技有限公司 | Cabinet type intelligent design method, device, equipment and medium based on representation learning |
| CN114462207A (en) * | 2022-01-07 | 2022-05-10 | 广州极点三维信息科技有限公司 | Matching method, system, equipment and medium for home decoration template |
-
2022
- 2022-12-28 CN CN202211693388.9A patent/CN115795627B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200203017A1 (en) * | 2017-06-09 | 2020-06-25 | Alerte Digital Sport Pty Ltd | Systems and methods of prediction of injury risk with a training regime |
| US20210064792A1 (en) * | 2019-08-27 | 2021-03-04 | Lg Electronics Inc. | Artificial intelligence device for guiding furniture placement and method of operating the same |
| CN113538101A (en) * | 2021-07-26 | 2021-10-22 | 杨浩松 | Method, system and storage medium for customizing home decoration based on artificial intelligence |
| CN113743009A (en) * | 2021-08-31 | 2021-12-03 | 广州极点三维信息科技有限公司 | Cabinet type intelligent design method, device, equipment and medium based on representation learning |
| CN114462207A (en) * | 2022-01-07 | 2022-05-10 | 广州极点三维信息科技有限公司 | Matching method, system, equipment and medium for home decoration template |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115795627B (en) | 2023-09-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7280039B2 (en) | Generation of 3D models representing buildings | |
| CN109145352B (en) | Data set for learning a function taking an image as input | |
| CN111381919B (en) | Forming a dataset for inferring an editable feature tree | |
| CN111382496B (en) | Learning neural networks for inferring editable feature trees | |
| CN109446413B (en) | Serialized recommendation method based on article association relation | |
| CN101675458B (en) | The automatic generation of the assembling instruction of building element model | |
| CN110060255A (en) | Semantic segmentation is carried out to 2D plan view using classifier pixel-by-pixel | |
| CN113767388A (en) | Techniques for workflow analysis and design task optimization | |
| CN105761303A (en) | Creation Of Bounding Boxes On 3d Modeled Assembly | |
| CN114140083A (en) | General data environment application system based on building information model | |
| CN112035588B (en) | Method for constructing spatial data rule base engine and GIS data quality inspection method | |
| JP2003132099A (en) | Method and apparatus for creating analytical model | |
| CN111382329A (en) | Data mining method and system for big data analysis | |
| Ritchie et al. | Neurosymbolic models for computer graphics | |
| KR20230082668A (en) | Computer Architecture for Creating Footwear Digital Assets | |
| US11281817B2 (en) | Systems and methods for generating programmatic designs of structures | |
| CN110737805A (en) | Method and device for processing graph model data and terminal equipment | |
| CN114417161B (en) | Virtual article time sequence recommendation method, device, medium and equipment based on special-purpose map | |
| CN114064926A (en) | Multi-modal power knowledge graph construction method, device, equipment and storage medium | |
| Lipp et al. | Local editing of procedural models | |
| CN116821223A (en) | Industrial visual control platform and method based on digital twinning | |
| US20200134909A1 (en) | Shaped-based techniques for exploring design spaces | |
| CN115795627A (en) | Furniture feature construction method, system, device and medium | |
| CN107169066A (en) | One kind is based on kdTree and the timing diagram data processing method of multivalued decision diagram | |
| US11126330B2 (en) | Shaped-based techniques for exploring design spaces |
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 |