Object arrangement design method, device, electronic device and storage medium

Abstract

The embodiments of the present application disclose an object arrangement design method, device, electronic device and storage medium. The method includes: obtaining feature information of the area to be designed, the feature information includes the name information of the area to be designed and the coordinate information of the area to be designed; inputting the feature information into an object arrangement model to obtain the object arrangement information output by the object arrangement model; obtaining the object pattern corresponding to the object arrangement information from an object template library to output the object arrangement design pattern corresponding to the object arrangement information. Through the above method, through the constructed object arrangement model, the object arrangement design scheme of the area to be designed can be automatically output, thereby improving the intelligence of the object arrangement design.

Description

Technical Field

[0001] The present application relates to the technical field of object layout design, and more specifically, to an object layout design method, device, electronic device and storage medium. Background Art

[0002] As the first step in object layout design, object arrangement design is an indispensable part in real estate building decoration design and game virtual reality set design. With the development of AI technology, the use of AI technology to quickly generate object arrangement design solutions is increasingly in line with market demand. However, the relevant object arrangement design methods require designers to use professional design software to provide users with object arrangement design solutions based on user needs and their own understanding of user needs, and the degree of intelligence is not high. Summary of the invention

[0003] In view of the above problems, the embodiments of the present application propose an object arrangement design method, device, electronic device and storage medium to improve the above problems.

[0004] In a first aspect, an embodiment of the present application provides an object arrangement design method, the method comprising: obtaining feature information of an area to be designed, the feature information comprising name information of the area to be designed and coordinate information of the area to be designed; inputting the feature information into an object arrangement model to obtain object arrangement information output by the object arrangement model; obtaining an object pattern corresponding to the object arrangement information from an object template library to output an object arrangement design pattern corresponding to the object arrangement information.

[0005] In a second aspect, an embodiment of the present application provides an object arrangement design device, the device comprising: an information acquisition unit, used to acquire feature information of an area to be designed, the feature information comprising name information of the area to be designed and coordinate information of the area to be designed; a first result output unit, used to input the feature information into an object arrangement model, and acquire object arrangement information output by the object arrangement model; a second result output unit, acquiring an object pattern corresponding to the object arrangement information from an object template library, to output a corresponding object arrangement design pattern.

[0006] In a third aspect, an embodiment of the present application provides an electronic device, comprising one or more processors and a memory; one or more programs are stored in the memory and configured to be executed by the one or more processors to implement the above method.

[0007] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which a program code is stored, wherein the above method is executed when the program code is executed by a processor.

[0008] The embodiments of the present application provide an object arrangement design method, device, electronic device, and storage medium, which obtains feature information of the area to be designed, wherein the feature information includes the name information of the area to be designed and the coordinate information of the area to be designed, inputs the feature information into a pre-trained object arrangement model, obtains the object arrangement information output by the object arrangement model, and then obtains the object pattern corresponding to the object arrangement information from the object template library to output the object arrangement design pattern corresponding to the object arrangement information. Through the above method, the object arrangement model constructed can automatically output the object arrangement design scheme of the area to be designed, thereby improving the intelligence of the object arrangement design. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A flow chart of an object arrangement design method proposed in one embodiment of the present application is shown;

[0011] Figure 2 A schematic diagram of coordinate information proposed in an embodiment of the present application is shown;

[0012] Figure 3 A flow chart of an object arrangement design method proposed in another embodiment of the present application is shown;

[0013] Figure 4 A schematic diagram of the structure of an object arrangement model proposed in another embodiment of the present application is shown;

[0014] Figure 5 A schematic diagram of calculating an object placement position vector according to another embodiment of the present application is shown;

[0015] Figure 6 A schematic diagram of calculating an object name vector proposed in another embodiment of the present application is shown;

[0016] Figure 7 A flowchart of an object arrangement design method proposed in another embodiment of the present application is shown;

[0017] Figure 8 A structural block diagram of an object arrangement design device proposed in one embodiment of the present application is shown;

[0018] Fig. 9 A structural block diagram of another object arrangement design device proposed in an embodiment of the present application is shown;

[0019] Fig.10 A structural block diagram of another object arrangement design device proposed in an embodiment of the present application is shown;

[0020] Fig.11 A structural block diagram of an electronic device used to execute the object arrangement design method according to an embodiment of the present application is shown in real time of the present application;

[0021] Fig.12 A storage unit in real time of the present application for storing or carrying program codes for implementing the object arrangement design method according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0023] As the first step in object layout design, object arrangement design is an indispensable part in real estate building decoration design and game virtual reality set design. With the development of AI technology, the use of AI technology to quickly generate object arrangement design solutions is increasingly in line with market demand.

[0024] However, the inventor found in the study of the relevant object arrangement design method that the relevant object arrangement design method requires designers to use professional design software to provide users with object arrangement design solutions based on user needs and their own understanding of user needs, and the degree of intelligence is not high. For example, in terms of housing decoration design, it is mainly the designer who uses professional design software to provide users with housing decoration design solutions based on user design needs and their own understanding of user needs. This design method may have the following problems: first, the communication cost is high, and the designer may not be able to accurately understand the user's real needs, and repeated communication and confirmation are required; second, the user's waiting time is long, and in addition to the time-consuming communication, the designer also needs time to provide housing decoration design solutions, and may need to modify the housing decoration design solutions many times. Due to the above reasons, the degree of intelligence of housing decoration design needs to be improved.

[0025] Therefore, the inventors proposed the method, device, electronic device and storage medium for object arrangement design in the embodiments of the present application, obtaining feature information of the area to be designed, the feature information including the name information of the area to be designed and the coordinate information of the area to be designed, inputting the feature information into a pre-trained object arrangement model, obtaining the object arrangement information output by the object arrangement model, and then obtaining the object pattern corresponding to the object arrangement information from the object template library to output the object arrangement design pattern corresponding to the object arrangement information. Through the above method, through the constructed object arrangement model, the object arrangement design scheme of the area to be designed can be automatically output, thereby improving the intelligence of the object arrangement design.

[0026] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0027] See also Figure 1 , an embodiment of the present application provides an object arrangement design method, the method comprising:

[0028] S110: Acquire characteristic information of the area to be designed, where the characteristic information includes name information of the area to be designed and coordinate information of the area to be designed.

[0029] As a method, the name information of the area to be designed is the name of each room in the area to be designed, such as the name of the bedroom, living room, kitchen, etc. The coordinate information of the area to be designed may include the outline information of the area to be designed. The coordinate information may be the coordinates of the intersection of the outline connection lines of the area to be designed, such as Figure 2 , you can Figure 2 Set the coordinates at the position shown in 1.

[0030] In an embodiment of the present application, before obtaining the characteristic information of the area to be designed, an image of the area to be designed may be obtained first, and then the characteristic information of the area to be designed may be extracted from the image of the area to be designed. Optionally, the image of the area to be designed may be obtained in a variety of ways. As one of the ways, the image of the area to be designed may be acquired in real time by an image acquisition device. Specifically, when the image of the area to be designed is acquired in real time by an image acquisition device, multiple images of the area to be designed may be acquired, and then the image with the highest clarity may be selected from the multiple images of the area to be designed as the final image of the area to be designed. As another way, a drawing of the area to be designed may also be directly input as the image of the area to be designed. Among them, the drawing of the area to be designed may be a design drawing generated by a specified design software.

[0031] After the image of the area to be designed is acquired by the above method, the name information and coordinate information of the area to be designed are extracted from the image of the area to be designed.

[0032] S120: Inputting the feature information into an object arrangement model, and acquiring object arrangement information output by the object arrangement model.

[0033] As a method, the object arrangement model is a pre-trained PGN (Pointer Generator Networks) network model, and the object arrangement model is used to output the name information and coordinate information of the object according to the input room name information and coordinate information of the area to be designed, so as to obtain the arrangement information of the object in the area to be designed, wherein the arrangement information can be the placement position, name and other information of the object.

[0034] Optionally, the object arrangement model may be a furniture arrangement model, which is used to output corresponding furniture arrangement information according to the input feature information of the area to be designed, wherein the furniture arrangement information represents the placement of furniture, for example, a bedside table is placed at the head of a bed in a bedroom.

[0035] S130: Acquire an object pattern corresponding to the object arrangement information from an object template library to output an object arrangement design pattern corresponding to the object arrangement information.

[0036] In an embodiment of the present application, the object template library may include a furniture template library, which stores a large number of furniture patterns of different styles, sizes, colors and materials.

[0037] The object arrangement design pattern is a visual object arrangement design pattern, and the user can view information such as the placement position, style, size, color, and material of the objects in the area to be designed.

[0038] Specifically, when the object arrangement information output by the object arrangement model is obtained, an object pattern corresponding to the object arrangement information can be searched from the object template library, and then an object arrangement design pattern of the area to be designed can be generated according to the corresponding object pattern found.

[0039] The present embodiment provides an object arrangement design method, which obtains feature information of a region to be designed, wherein the feature information includes name information of the region to be designed and coordinate information of the region to be designed, inputs the feature information into a pre-trained object arrangement model, obtains object arrangement information output by the object arrangement model, and then obtains an object pattern corresponding to the object arrangement information from an object template library to output an object arrangement design pattern corresponding to the object arrangement information. Through the above method, the object arrangement model constructed can automatically output the object arrangement design scheme of the region to be designed, thereby improving the intelligence of the object arrangement design.

[0040] See also Figure 3 , an embodiment of the present application provides an object arrangement design method, the method comprising:

[0041] S210: Establish a first network model.

[0042] As a method, the first network model is an initial object arrangement model, and the initial object arrangement model may be a furniture arrangement model. The parameters of the first network model may be initialized in advance, and initial values ​​may be set for the parameters of the first network model.

[0043] S220: Obtain the labeled object arrangement design drawing data.

[0044] In the embodiment of the present application, after the first network model is established, the first network model needs to be trained, so a large amount of training data needs to be obtained to train the first network model.

[0045] The object arrangement design drawing data may be furniture arrangement design drawing data.

[0046] When acquiring training data, a large number of furniture arrangement design drawings can be acquired from a furniture arrangement design library. The furniture arrangement design library is a database for storing existing furniture arrangement design drawings. After acquiring a large number of furniture arrangement design drawings, the furniture arrangement design drawings can be sorted and screened, and then the furniture arrangement design drawings can be annotated with data using an annotation tool to obtain the annotated furniture arrangement design drawing data. Optionally, when performing data annotation, the data annotation can be divided into two parts, the first part is to annotate the room information in the furniture arrangement design drawings, and the second part is to annotate the furniture information in the furniture arrangement design drawings. The room information includes the room name information and the room coordinate information, and the furniture information includes the furniture name and the furniture coordinate information.

[0047] When data annotation is performed on the furniture arrangement design drawing, the room information and furniture information are annotated on each wall in the furniture arrangement design drawing in a certain order, and then the annotated room information and furniture information are processed into an input sequence and an output sequence of the first network model. The certain order may be a clockwise order or a counterclockwise order, which is not specifically limited here. The input sequence includes a room name character sequence and a room coordinate sequence, and the output sequence includes a furniture name character sequence and a furniture coordinate sequence.

[0048] S230: Processing the annotated object arrangement design drawing data to obtain room information and object information corresponding to the object arrangement design drawing data, wherein the room information includes a room name character sequence and a room coordinate sequence, and the object information includes an object name character sequence and an object coordinate sequence.

[0049] As a method, the obtained annotated furniture arrangement design drawing data is preprocessed. The preprocessing may include: first, performing unified scale processing on the coordinate information of the room and the coordinate information of the furniture. Specifically, a scaling factor is calculated for the scales of different furniture arrangement design drawings, and all furniture arrangement design drawing data are scaled uniformly according to the calculated scaling factor; second, adding a start identifier ( <sos>), add end identifiers to the end of the input and output sequences ( <eos>), specifically, add a start identifier ( <sos>), add an end marker ( <eos>); Add a start identifier to the beginning of the furniture name character sequence and the furniture coordinate sequence ( <sos>), add an end identifier ( <eos>); Third, unify the input sequence and the output sequence according to the maximum length. If the length of the input sequence or the output sequence is not the maximum length, fill the end of the input sequence or the output sequence with 0 to make it the maximum length.

[0050] After the above operations, the pre-processed room information and furniture information corresponding to the furniture arrangement design drawing data are obtained.

[0051] S240: Using the room information as an input sequence of the first network model, obtaining the output of the first network model as the object information, training the first network model, and obtaining an object arrangement model.

[0052] In an embodiment of the present application, the object arrangement model includes a first encoding network, a first decoding network and a second decoding network; wherein the first encoding network is used to receive feature information of the area to be designed, and encode the feature information, and input the encoding vectors obtained after encoding into the first decoding network and the second decoding network respectively; the first decoding network is used to decode the coordinate sequence in the vector information to obtain the placement position information of the objects in the object arrangement sequence, and the second decoding network is used to decode the name character sequence in the vector information to obtain the name information of the objects in the object arrangement sequence.

[0053] Specifically, Figure 4 The structural diagram of a furniture arrangement model shown in the figure, the furniture arrangement model 200 includes an encoding network 210 and a first decoding network 220 and a second decoding network 230. The first encoding network 210 and the first decoding network 220 and the second decoding network 230 are all LSTM (Long Short-Term Memory), which is a time recursive neural network suitable for processing and predicting important events with relatively long intervals and delays in time series. Among them, the first encoding network 210 is used to receive the room name information (such as bedroom, living room, kitchen, etc.) of the area to be designed and the coordinate information (such as the coordinate information of the bedroom, the coordinate information of the living room, etc.) of the area to be designed for encoding, and input the encoded code vectors into the first decoding network 220 and the second decoding network 230 respectively. The encoded code vectors correspond to the room name character sequence and the coordinate sequence respectively. Among them,

[0054] The first decoding network 220 is used to decode the room name character sequence in the coding vector to obtain the name information of the furniture in the area to be designed. The second decoding network 230 is used to decode the coordinate sequence in the coding vector and perform furniture arrangement design based on the decoding result to obtain the placement position information of the furniture in the area to be designed.

[0055] By designing two decoding networks, the decoding ability of the furniture arrangement model can be improved, and the furniture position coordinate sequence and the furniture name character sequence can be accurately decoded, thereby improving the furniture arrangement design ability of the entire furniture arrangement design model.

[0056] Among them, the step of obtaining the first decoding network includes: obtaining a first target loss function, the first target loss function includes a loss function calculated based on the first attention weight; training the first decoding network based on the first target loss function to obtain a trained first decoding network.

[0057] Specifically, for the first decoding network, when training the first decoding network, a coverage mechanism is introduced into the loss function part of the first decoding network. The coverage mechanism is mainly used to alleviate the problem of many repeated characters appearing in decoding. The specific formula of the coverage mechanism is as follows:

[0058]

[0059]

[0060] Among them, c t is the accumulation of the first attention weight at time t-1, a t is the first attention weight, and the calculation formula of the first attention weight is as follows:

[0061]

[0062] a t =softmax(e t )------------(4); where v, W h ,W s ,b are parameters that need to be trained, h i is the implicit state vector of the first encoding network at time t, s t is the implicit state vector of the first decoding network and the second decoding network at time t.

[0063] The first attention weight at the current moment and c t For comparison, add this as loss to the first objective loss function, and the final first objective loss function is: Among them, θ is a parameter, is a pre-set probability parameter.

[0064] The first decoding network is iteratively trained using the above loss function. When it is detected that the value of the loss function is lower than a preset value, or the number of iteration steps reaches a preset number of iteration steps, the training of the first decoding network is stopped to obtain the final first decoding network.

[0065] The step of obtaining the second decoding network includes: obtaining a second target loss function, the second target loss function includes a wall distance loss function and an object position overlap loss function, wherein the wall distance loss function is a loss function corresponding to the wall distance generated according to the prior distance between the object and the wall; the object position overlap loss function is a loss function corresponding to the overlap area generated according to the overlap area between objects, the overlap area between the object and the door, the area of ​​the object beyond the room, and the overlap area between the object and the window; the second decoding network is trained based on the second target loss function to obtain the trained second decoding network.

[0066] Among them, the object position overlap loss function is specifically a furniture position overlap loss function.

[0067] Specifically, for the second decoding network, a penalty for overlapping the wall distance and the furniture position is added to the loss function. These penalties are added to make the arrangement of the furniture more reasonable. In the embodiment of the present application, the wall distance loss function is:

[0068] in, Among them, λ is a parameter, d i is the distance between the furniture and the wall, It is an empirical statistical value, which is the reasonable distance from the wall. i <0, the value of λ is greater than 1, which means that when the furniture exceeds the wall, the loss function value increases rapidly; when When , the value of λ is greater than or equal to 0 and less than 0.1, which means that when the furniture exceeds the wall, the loss function value is very small; when , the value of λ is greater than 0.5 and less than 1, which means that when the furniture goes beyond the wall, the loss function value grows slowly.

[0069] The furniture position overlap loss function is: Among them, f i , r i ,s i , w i is the weight parameter, F i is the overlapping area between furniture, R i is the overlapping area between the furniture and the door, S i W is the area of ​​furniture that extends beyond the room. i It is the overlapping area between furniture and windows.

[0070] Finally, the loss function of the second decoding network is: Among them, delt(σ) is the pre-set loss parameter.

[0071] The second decoding network is iteratively trained using the above loss function. When it is detected that the value of the loss function is lower than a preset value, or the number of iteration steps reaches a preset number of iteration steps, the training of the second decoding network is stopped to obtain the final second decoding network.

[0072] Optionally, in an embodiment of the present application, the placement position information of the object decoded by the first decoding network includes an object placement vector, and the step of obtaining the object placement vector includes: obtaining a first weight, wherein the first weight is the weight of the object placement position calculated based on the first decoding vector and the encoding vector, and the first decoding vector is a decoding vector obtained after decoding by the first decoding network; performing weighted summation of the first weight and the encoding vector to obtain a first attention weight vector; and concatenating the first attention weight vector with the decoding vector to obtain the object placement position vector.

[0073] pass Figure 5 To illustrate the above, Figure 5 As shown, in the case of four furniture placement positions, the furniture placement position decoding vectors 13 corresponding to the four furniture placement positions obtained by the first eigenvector representation include B1, B2, B3 and B4. Among them, B1, B2, B3 and B4 correspond to one furniture placement position decoding vector in the four furniture placement decoding vectors. The furniture placement position encoding vectors 14 corresponding to the four furniture placement positions obtained by the first eigenvector representation include B5, B6, B7 and B8. Among them, B5, B6, B7 and B8 correspond to one furniture placement position encoding vector in the four furniture placement position encoding vectors.

[0074] If the first weight calculation is currently being performed for B1, then B1 and B5 can be weighted to obtain the weight a1 between B1 and B5, and then the corresponding weight a2 between B1 and B6, the weight a3 between B1 and B7, and the weight a4 between B1 and B8 can be calculated. Then, the weight a1, weight a2, weight a3, and weight a4 can all be used as the first weight. Then, a weighted sum calculation is performed, that is, (B5xa1)+(B6xa2)+(B7xa3)+(B8xa4)=M1 is calculated, and M1 is used as the first attention weight vector. Then, M1 is spliced ​​with B1 to obtain the furniture placement vector of the furniture corresponding to B1. By analogy, the furniture placement vector of the furniture corresponding to B2 can be obtained, the furniture placement vector of the furniture corresponding to B3 can be obtained, and the furniture placement vector of the furniture corresponding to B4 can be obtained, and then a new furniture placement vector 15 can be obtained.

[0075] It should be noted that in this embodiment, the representation of the furniture placement position is based on the vector form. When calculating the first weight, it can be understood as calculating the weight between two vectors. When calculating the weight between two vectors, the corresponding elements between the two vectors can be multiplied to obtain multiple products, and then the value obtained by adding the multiple products is used as the weight between the two vectors, and then as the weight of the object placement position. Among them, the corresponding elements can be understood as the elements with the same position in the respective vector representations. For example, if B1 = [0, 1, 2, 3], B5 = [1, 2, 3, 3]. Then the element corresponding to the first element "0" in B1 is the first element "1" in B5, the element corresponding to the second element "1" in B1 is the second element "2" in B5, the element corresponding to the third element "2" in B1 is the third element "3" in B5, and the element corresponding to the fourth element "3" in B1 is the fourth element "3" in B5. The first weight calculated is 0x1+1x2+2x3+3x3=17.

[0076] It should also be noted that in the process of splicing two vector representations, it can be understood that the elements included in each of the two vector representations are all used as elements in the spliced ​​vector representation. For example, if one vector representation is [1,0,2,1] and the other vector representation is [1,0,1], then the spliced ​​vector representation is [1,0,2,1,1,0,1]. It can be understood that the first four elements in the spliced ​​vector representation originally belong to one vector representation, and the last three elements in the spliced ​​vector representation originally belong to another vector representation.

[0077] The name information of the object decoded by the second decoding network includes an object name vector, and the step of obtaining the object name vector includes: obtaining a second weight, wherein the second weight is the weight of the object name calculated based on the second decoding vector and the encoding vector, and the second decoding vector is a decoding vector obtained after decoding by the second decoding network; weighted summing the second weight and the encoding vector to obtain a second attention weight vector; and concatenating the second attention weight vector and the encoding vector to obtain the object name vector.

[0078] Next, pass Figure 6 The above contents are illustrated with examples.

[0079] like Figure 6 As shown, in the case of four furniture names, the contents of the furniture name decoding vector 16 and the furniture name encoding vector 17 corresponding to the four furniture names obtained by the first feature vector are the same as Figure 4 The content shown in is the same. If B5 is currently performing the second weight calculation, then B5 and B1 can be calculated to obtain the weight c1 between B5 and B1, and the corresponding weight c2 between B5 and B2, the weight c3k between B5 and B3, and the weight c4 between B5 and B4 can be calculated. Then the weight c1, weight c2, weight c3 and weight c4 can be used as the second weight. Then perform a weighted summation calculation, that is, calculate (B1x c1)+(B2xc2)+(B3xc3)+(B4xc4)=M2, and use M2 as the second attention weight vector, and then concatenate N2 with B5 to obtain the furniture name vector of the furniture corresponding to B5. By analogy, the furniture name vector of the furniture corresponding to B6 can be obtained, the furniture name vector of the furniture corresponding to B7 can be obtained, and the furniture name vector of the furniture corresponding to B8 can be obtained, and then a new furniture name vector 18 can be obtained.

[0080] Among them, the principle of calculating the second weight in the above content is the same as the principle of calculating the first weight mentioned above, and the principle of splicing the second attention weight vector with the corresponding furniture name coding vector is also the same as the principle of splicing each first attention weight vector with the corresponding furniture placement position coding vector mentioned above, which will not be repeated here.

[0081] S250: Acquire characteristic information of the area to be designed, where the characteristic information includes name information of the area to be designed and coordinate information of the area to be designed.

[0082] S260: Input the feature information into the object arrangement model to obtain the object arrangement information output by the object arrangement model.

[0083] The detailed explanation of the steps included in S250 and S260 can refer to the corresponding steps in the aforementioned embodiment, which will not be repeated here.

[0084] S270: Inputting the object arrangement information and the object template library into a deep semantic matching model to obtain an object arrangement design pattern output by the deep semantic matching model.

[0085] As a method, the deep semantic matching model is a DSSM network model (Deep Structured Semantic Models). The structural diagram of the deep semantic matching model is as follows: Figure 8 As shown, it includes input layer, representation layer and matching layer.

[0086] If the object arrangement information is furniture arrangement information, then the furniture name information and the furniture template library in the furniture arrangement information obtained by the above method can be input into the deep semantic matching model, and the feature information is extracted through the CNN convolutional network of the representation layer, and then the matching degree is calculated in the matching layer, and finally the corresponding furniture arrangement design pattern is output.

[0087] S280: If the deep semantic matching model outputs multiple sets of object arrangement design patterns, determine the target object arrangement design pattern of the area to be designed from the multiple sets of object arrangement design patterns according to target requirements.

[0088] As a method, if the furniture template library stores multiple sets of furniture patterns that match the input furniture name information, the deep semantic matching model can output multiple sets of furniture arrangement design patterns according to the input furniture name information. Through the above method, the diversity of generated furniture arrangement design solutions can be increased.

[0089] In the embodiment of the present application, the target requirement is information input by the user, which may be the size, style, material, price, etc. of the furniture. When the furniture arrangement design patterns output by the deep semantic matching model are multiple sets, the target furniture arrangement design pattern of the area to be designed can be determined from the multiple sets of furniture arrangement design patterns according to the information such as the size, style, material, and price of the furniture input by the user.

[0090] The present embodiment provides an object arrangement design method, firstly establishes a first network model, obtains the annotated object arrangement design drawing data, processes the annotated object arrangement design drawing data, obtains the room information and object information corresponding to the object arrangement design drawing data, uses the room information as the input sequence of the first network model, obtains the output of the first network model as the object information, trains the first network model to obtain the object arrangement model, then obtains the feature information of the area to be designed, inputs the feature information into the object arrangement model, obtains the object arrangement information output by the object arrangement model, then inputs the object arrangement information and the object template library into the deep semantic matching model, obtains the object arrangement design pattern output by the deep semantic matching model, and if the object arrangement design pattern output by the deep semantic matching model is multiple sets, according to the target requirements, determines the target object arrangement design pattern of the area to be designed from the multiple sets of object arrangement design patterns. Through the above method, through the constructed object arrangement model and deep semantic matching model, the object arrangement design scheme of the area to be designed can be automatically output, thereby improving the intelligence of the object arrangement design.

[0091] See also Figure 7 , an embodiment of the present application provides an object arrangement design method, the method comprising:

[0092] S310: Acquire characteristic information of the area to be designed, where the characteristic information includes name information of the area to be designed and coordinate information of the area to be designed.

[0093] S320: Input the feature information into an object arrangement model, and obtain object arrangement information output by the object arrangement model, wherein the object arrangement model is used to output corresponding object arrangement information according to the input feature information of the area to be designed.

[0094] S330: Inputting the object arrangement information and the object template library into a deep semantic matching model to obtain an object arrangement design pattern output by the deep semantic matching model.

[0095] The detailed explanation of the steps included in S310, S320 and S330 can refer to the corresponding steps in the aforementioned embodiments, which will not be repeated here.

[0096] S340: If the object arrangement design patterns output by the deep semantic matching model are multiple sets, obtain attribute information of the object models in the object arrangement design library and a matching relationship between the first object and the second object.

[0097] As a method, the attribute information of the object may include the style, size, material and price of the object. The object arrangement design library is a database, which stores a large number of object arrangement design schemes and a large number of object models.

[0098] When there are multiple sets of object arrangement design patterns output by the deep semantic model, one set can be selected as the final object arrangement design pattern. The final object arrangement design pattern can be determined from multiple sets of object arrangement design patterns by the following method. Specifically, a knowledge graph library of objects can be pre-built, and then the knowledge graph library of objects can be queried to determine the final object arrangement design pattern from multiple sets of object arrangement design patterns.

[0099] In the embodiment of the present application, when it is determined that the object arrangement design pattern output by the deep semantic matching model is a plurality of sets, the attribute information of the object model (the style, size, material and price of the object, etc.) and the matching relationship between the first object and the first object, etc. are obtained from the object arrangement design library. The matching relationship between the first object and the first object is the matching relationship between different objects in the object arrangement design library, and the matching relationship between the first object and the second object can be pre-stored in the object arrangement design library, and can then be directly obtained from the object arrangement design library.

[0100] S350: Constructing a knowledge graph library of the object based on the attribute information and the matching relationship between the first object and the second object.

[0101] As a method, before constructing the knowledge graph library of the object, it is necessary to obtain the attributes of the object model and the matching relationship between the first object and the second object from the object arrangement design library, and then the knowledge graph library of the object can be established based on the acquired data. For example, if the object is furniture, then the attribute information of the furniture model (such as the style, size, material, price, etc. of the furniture) and the matching relationship between the first furniture and the second furniture (such as the matching relationship between the bed and the bedside table) can be obtained from the furniture arrangement design library, and then the knowledge graph library of the furniture can be established based on the above-obtained furniture attribute information and the matching relationship between the first furniture and the second furniture.

[0102] S360: According to the target requirement, search the knowledge graph library of the object, and determine the target object arrangement design pattern of the area to be designed from multiple sets of the object arrangement design patterns.

[0103] As a method, the target requirement is information input by the user, which may be the size, style, material, price, etc. of the object. When the user has the object knowledge to replace a set of object arrangement design patterns among multiple sets of object arrangement design patterns, the user can input the size, style, material, price, etc. of the object, search from the constructed object knowledge graph library, and then replace the object with a different style or material, and then use the object arrangement design pattern after the object replacement as the target object arrangement design pattern.

[0104] The present embodiment provides an object arrangement design method, firstly, obtaining feature information of the area to be designed, wherein the feature information includes the name information of the area to be designed and the coordinate information of the area to be designed, then inputting the feature information into the object arrangement model, obtaining the object arrangement information output by the object arrangement model, then inputting the object arrangement information and the object template library into the deep semantic matching model, obtaining the object arrangement design pattern output by the deep semantic matching model, if there are multiple sets of object arrangement design patterns output by the deep semantic matching model, obtaining the attribute information of the object model in the object arrangement design library and the matching relationship between the first object and the second object, constructing the object knowledge graph library based on the attribute information and the matching relationship between the first object and the second object, and finally searching the object knowledge graph library according to the target requirements, and determining the target object arrangement design pattern of the area to be designed from the multiple sets of object arrangement design patterns. Through the above method, the knowledge graph technology is introduced, and the object knowledge graph library is constructed, so that the automatic replacement of the object can be realized according to the target requirements, thereby improving the diversity of the generated object arrangement design patterns.

[0105] See also Figure 8 The present application provides an object arrangement design device 400, the device 400 comprising:

[0106] The information acquisition unit 410 is used to acquire characteristic information of the area to be designed, where the characteristic information includes name information of the area to be designed and coordinate information of the area to be designed.

[0107] The first result output unit 420 is used to input the feature information into the object arrangement model to obtain the object arrangement information output by the object arrangement model.

[0108] Wherein, the object arrangement model includes a first encoding network, a first decoding network and a second decoding network; wherein, the first encoding network is used to receive feature information of the area to be designed, and encode the feature information, and input the encoded vectors obtained after encoding into the first decoding network and the second decoding network respectively; the first decoding network is used to decode the coordinate sequence in the vector information to obtain the placement position information of the objects in the object arrangement sequence, and the second decoding network is used to decode the name character sequence in the vector information to obtain the name information of the objects in the object arrangement sequence.

[0109] The second result output unit 430 obtains the object pattern corresponding to the object arrangement information from the object template library to output the corresponding object arrangement design pattern.

[0110] Optionally, the second result output unit 430 is used to input the object arrangement information and the object template library into the deep semantic matching model to obtain the object arrangement design pattern output by the deep semantic matching model.

[0111] Please refer to 9, the device 400 further includes:

[0112] The determination unit 440 is configured to determine a target object arrangement design pattern of the area to be designed from the multiple sets of object arrangement design patterns according to target requirements if the deep semantic matching model outputs multiple sets of object arrangement design patterns.

[0113] Specifically, the determination unit 440 obtains the attribute information of the object model in the object arrangement design library and the matching relationship between the first object and the second object; constructs a knowledge graph library of the object based on the attribute information and the matching relationship between the first object and the second object; searches the knowledge graph library of the object according to the target requirements, and determines the target object arrangement design pattern of the area to be designed from multiple sets of the object arrangement design patterns.

[0114] See also Fig.10 , the device 400 further includes:

[0115] The network training unit 450 is used to obtain a first weight, wherein the first weight is the weight of the object placement position calculated based on the first decoding vector and the encoding vector, and the first decoding vector is a decoding vector obtained after decoding by the first decoding network; the first weight is weighted and summed with the encoding vector to obtain a first attention weight vector; the first attention weight vector is concatenated with the decoding vector to obtain the object placement position vector.

[0116] Specifically, the network training unit 450 is used to obtain a first target loss function, which includes a loss function calculated based on the first attention weight; the first decoding network is trained based on the first target loss function to obtain a trained first decoding network.

[0117] Optionally, the network training unit 450 is also used to obtain a second weight, wherein the second weight is the weight of the object name calculated based on the second decoding vector and the encoding vector, and the second decoding vector is a decoding vector obtained after decoding by the second decoding network; the second weight is weightedly summed with the encoding vector to obtain a second attention mechanism weight vector; the second attention mechanism weight vector is concatenated with the encoding vector to obtain the object name vector.

[0118] Specifically, the network training unit 450 is used to obtain a second target loss function, which includes a wall distance loss function and an object position overlap loss function, wherein the wall distance loss function is a loss function corresponding to the wall distance generated based on a priori distance between the object and the wall; the object position overlap loss function is a loss function corresponding to the overlapping area generated based on the overlapping area between objects, the overlapping area between the object and the door, the area of ​​the object beyond the room, and the overlapping area between the object and the window; the second decoding network is trained based on the second target loss function to obtain a trained second decoding network.

[0119] It should be noted that the device embodiment in the present application corresponds to the aforementioned method embodiment. The specific principles in the device embodiment can be found in the contents of the aforementioned method embodiment and will not be repeated here.

[0120] The following will be combined Fig.11 An electronic device provided by the present application is described.

[0121] See also Fig.11 Based on the above-mentioned object arrangement design method and device, the embodiment of the present application also provides another electronic device 100 that can execute the above-mentioned object arrangement design method. The electronic device 100 includes one or more (only one is shown in the figure) processors 102, a memory 104 and a network module 106 that are coupled to each other. Among them, the memory 104 stores a program that can execute the content of the above-mentioned embodiment, and the processor 102 can execute the program stored in the memory 104.

[0122] Among them, the processor 102 may include one or more processing cores. The processor 102 uses various interfaces and lines to connect various parts of the entire electronic device 100, and executes various functions and processes data of the electronic device 100 by running or executing instructions, programs, code sets or instruction sets stored in the memory 104, and calling data stored in the memory 104. Optionally, the processor 102 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 102 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing display content; and the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 102, but may be implemented separately through a communication chip.

[0123] The memory 104 may include a random access memory (RAM) or a read-only memory (ROM). The memory 104 may be used to store instructions, programs, codes, code sets or instruction sets. The memory 104 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the following various method embodiments, etc. The data storage area may also store data (such as a phone book, audio and video data, chat record data) created by the electronic device 100 during use.

[0124] The network module 106 is used to receive and send electromagnetic waves, realize the mutual conversion between electromagnetic waves and electrical signals, so as to communicate with a communication network or other devices, such as communicating with an audio playback device. The network module 106 may include various existing circuit elements for performing these functions, such as antennas, radio frequency transceivers, digital signal processors, encryption / decryption chips, user identity modules (SIM) cards, memories, etc. The network module 106 may communicate with various networks such as the Internet, corporate intranets, wireless networks, or communicate with other devices via wireless networks. The above-mentioned wireless networks may include cellular telephone networks, wireless local area networks, or metropolitan area networks. For example, the network module 106 may exchange information with a base station.

[0125] Please refer to Fig.12 , which shows a structural block diagram of a computer-readable storage medium provided in an embodiment of the present application. The computer-readable medium 800 stores program codes, which can be called by a processor to execute the method described in the above method embodiment.

[0126] The computer readable storage medium 800 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk, or a ROM. Optionally, the computer readable storage medium 800 includes a non-transitory computer-readable storage medium. The computer readable storage medium 800 has storage space for program code 810 that performs any method step in the above method. These program codes can be read from or written to one or more computer program products. The program code 810 can be compressed, for example, in an appropriate form.

[0127] The present embodiment provides an object arrangement design method, device, electronic device and storage medium, which obtains feature information of the area to be designed, wherein the feature information includes the name information of the area to be designed and the coordinate information of the area to be designed, inputs the feature information into a pre-trained object arrangement model, obtains the object arrangement information output by the object arrangement model, and then obtains the object pattern corresponding to the object arrangement information from the object template library to output the object arrangement design pattern corresponding to the object arrangement information. Through the above method, the object arrangement model constructed can automatically output the object arrangement design scheme of the area to be designed, thereby improving the intelligence of the object arrangement design.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.< / eos> < / sos> < / eos> < / sos> < / eos> < / sos>

Claims

1. A method for designing an object arrangement, characterized in that: The method comprises: Acquire characteristic information of the area to be designed, wherein the characteristic information includes name information of the area to be designed and coordinate information of the area to be designed; Inputting the feature information into an object arrangement model to obtain object arrangement information output by the object arrangement model; Acquire an object pattern corresponding to the object arrangement information from an object template library to output an object arrangement design pattern corresponding to the object arrangement information; Wherein, the object arrangement model includes a first encoding network, a first decoding network and a second decoding network; The first encoding network is used to receive the feature information of the area to be designed, encode the feature information, and input the encoding vector obtained after encoding into the first decoding network and the second decoding network respectively; The first decoding network is used to decode the coordinate sequence in the vector information to obtain the placement position information of the objects in the object arrangement sequence, and the second decoding network is used to decode the name character sequence in the vector information to obtain the name information of the objects in the object arrangement sequence; Wherein, the placement position information of the object includes the object placement position vector, and the method further includes: Obtaining a first weight, wherein the first weight is a weight of the object placement position calculated according to the first decoding vector and the encoding vector, and the first decoding vector is a decoding vector obtained after being decoded by the first decoding network; Performing a weighted summation of the first weight and the encoding vector to obtain a first attention weight vector; The first attention weight vector is concatenated with the decoding vector to obtain the object placement position vector.

2. The method according to claim 1, characterized in that: The acquiring the object pattern corresponding to the object arrangement information from the object template library to output the corresponding object arrangement design pattern includes: The object arrangement information and the object template library are input into a deep semantic matching model to obtain an object arrangement design pattern output by the deep semantic matching model.

3. The method according to claim 2, characterized in that After inputting the object arrangement sequence and the object template library sequence into the deep semantic matching model and obtaining the object arrangement design pattern output by the deep semantic matching model, the method further includes: If the deep semantic matching model outputs multiple sets of object arrangement design patterns, the target object arrangement design pattern of the area to be designed is determined from the multiple sets of object arrangement design patterns according to target requirements.

4. The method according to claim 3, characterized in that The step of determining the target object arrangement design pattern of the area to be designed from the plurality of sets of the object arrangement design patterns according to the target requirement comprises: Acquire attribute information of object models in the object arrangement design library and a matching relationship between the first object and the second object; Building a knowledge graph library of objects based on the attribute information and the matching relationship between the first object and the second object; According to the target requirements, the knowledge graph library of the object is searched, and the target object arrangement design pattern of the area to be designed is determined from multiple sets of the object arrangement design patterns.

5. The method according to claim 1, characterized in that The method further comprises: Obtaining a first objective loss function, wherein the first objective loss function includes a loss function calculated based on the first attention weight; The first decoding network is trained based on the first objective loss function to obtain a trained first decoding network.

6. The method according to claim 1, characterized in that The name information of the object includes an object name vector, and the method further includes: Obtaining a second weight, wherein the second weight is a weight of the object name calculated according to a second decoding vector and the encoding vector, and the second decoding vector is a decoding vector obtained after being decoded by the second decoding network; Performing a weighted summation of the second weight and the encoding vector to obtain a second attention mechanism weight vector; The second attention mechanism weight vector is concatenated with the encoding vector to obtain the object name vector.

7. The method according to claim 6, characterized in that The method further comprises: Obtaining a second objective loss function, the second objective loss function includes a wall distance loss function and an object position overlap loss function, wherein the wall distance loss function is a loss function corresponding to the wall distance generated according to a priori distance between the object and the wall; the object position overlap loss function is a loss function corresponding to the overlap area generated according to the overlap area between objects, the overlap area between the object and the door, the area of ​​the object beyond the room, and the overlap area between the object and the window; The second decoding network is trained based on the second objective loss function to obtain a trained second decoding network.

8. An object arrangement design device, characterized in that: The device comprises: An information acquisition unit, used to acquire characteristic information of the area to be designed, wherein the characteristic information includes name information of the area to be designed and coordinate information of the area to be designed; A first result output unit, used for inputting the feature information into an object arrangement model to obtain object arrangement information output by the object arrangement model; A second result output unit, which obtains an object pattern corresponding to the object arrangement information from an object template library to output a corresponding object arrangement design pattern; Wherein, the object arrangement model includes a first encoding network, a first decoding network and a second decoding network; The first encoding network is used to receive the feature information of the area to be designed, encode the feature information, and input the encoding vector obtained after encoding into the first decoding network and the second decoding network respectively; The first decoding network is used to decode the coordinate sequence in the vector information to obtain the placement position information of the objects in the object arrangement sequence, and the second decoding network is used to decode the name character sequence in the vector information to obtain the name information of the objects in the object arrangement sequence; Among them, the placement position information of the object includes an object placement position vector, and a network training unit is used to obtain a first weight, wherein the first weight is the weight of the object placement position calculated based on a first decoding vector and the encoding vector, and the first decoding vector is a decoding vector obtained after decoding by the first decoding network; the first weight is weighted and summed with the encoding vector to obtain a first attention weight vector; the first attention weight vector is spliced ​​with the decoding vector to obtain the object placement position vector.

9. An electronic device, characterized in that: The method comprises one or more processors and a memory; one or more programs are stored in the memory and are configured so that the one or more processors execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program code, wherein when the program code is executed by a processor, the method according to any one of claims 1 to 7 is executed.

Images