CN110751721B - Furniture layout drawing generation method and device, computer equipment and storage medium - Google Patents
Furniture layout drawing generation method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN110751721B CN110751721B CN201911345314.4A CN201911345314A CN110751721B CN 110751721 B CN110751721 B CN 110751721B CN 201911345314 A CN201911345314 A CN 201911345314A CN 110751721 B CN110751721 B CN 110751721B
- Authority
- CN
- China
- Prior art keywords
- furniture
- information
- arrangement
- room
- house
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The application provides a furniture layout drawing generation method, a furniture layout drawing generation device, computer equipment and a computer readable storage medium, wherein the method comprises the following steps: acquiring a house drawing of a room; reading the house drawing by using a CNN network, extracting features, and acquiring quantitative data of the house; the CNN network is obtained based on collected decoration drawings through training; inputting the quantized data into an RNN sequence model for sequence arrangement, and outputting arrangement vectors of furniture; the RNN sequence model is obtained by training furniture arrangement information based on the decoration drawing; and searching furniture information of the corresponding furniture from a furniture library storing the decoration drawing according to the arrangement vector, and generating a furniture arrangement drawing of the room according to the furniture information. According to the scheme, the efficiency of generating the furniture layout drawing can be improved, and the use effect of the furniture layout drawing is better.
Description
Technical Field
The application relates to the technical field of furniture layout design, in particular to a furniture layout drawing generation method and device, computer equipment and a computer readable storage medium.
Background
The furniture arrangement design is used as a first ring of furniture layout design, and is an indispensable link in the decoration design of real estate buildings and the design of game virtual reality scenery.
With the rapid development of the internet, home layout systems and related solutions appear in the market. For a home layout system, taking Ku Jia le as an example, a plurality of furniture models which can be freely collocated are provided in a mode that a user draws a house type picture on line, the user can adjust the size and the position of a furniture 3D model at multiple angles, and finally a picture is rendered through a three-dimensional engine.
The scheme is time-consuming and labor-consuming, and the rendered furniture layout is often unable to easily buy furniture with the same size in the market due to the fact that the furniture model is stretched in different degrees; and the related method generally needs to generate the furniture sequence data set by manually marking data, so that the marking workload is large, the efficiency is low, and the constraint parameters are introduced to participate in the furniture size prediction, so that the work is various and the processing process is complex.
In summary, the existing furniture arrangement design scheme has the defects of low efficiency and complex treatment process.
Disclosure of Invention
The purpose of the present application is to solve at least one of the above technical drawbacks, in particular, the drawbacks of low efficiency and complicated processing procedure, and to provide a method and an apparatus for generating a furniture layout drawing.
A method for generating a furniture layout drawing comprises the following steps:
acquiring a house drawing of a room;
reading the house drawing by using a CNN network, extracting features, and acquiring quantitative data of the house; the CNN network is obtained based on collected decoration drawings through training; inputting the quantized data into an RNN sequence model for sequence arrangement, and outputting arrangement vectors of furniture; the RNN sequence model is obtained by training furniture arrangement information based on the decoration drawing;
and searching furniture information of the corresponding furniture from a furniture library storing the decoration drawing according to the arrangement vector, and generating a furniture arrangement drawing of the room according to the furniture information.
In an embodiment, the method for generating a furniture layout drawing further includes: collecting decoration drawings of existing rooms, wherein the decoration drawings comprise house information and furniture information corresponding to the house information;
acquiring three-dimensional data and position information of the ground furniture of the decoration drawing;
and establishing the furniture library according to the three-dimensional data and the position information thereof.
In one embodiment, the method for generating a furniture layout drawing further includes: acquiring the collected decoration drawings;
deleting the furniture information in the decoration drawing to obtain a picture format drawing containing the house information;
and training the CNN network by taking the picture format drawing as input.
In an embodiment, the method for generating a furniture layout drawing, where the drawing in picture format is used as an input to train a CNN network, includes: inputting the picture format drawing into a CNN network;
and performing feature extraction on the picture format drawing through a plurality of convolution cores of the CNN network, and converting the extracted house information into an intermediate state through nonlinear transformation.
In an embodiment, in the method for generating a furniture layout drawing, the room is a blank room; the acquiring of the house drawing of the room comprises the following steps:
acquiring blank room information input by a user, and converting the blank room information to obtain a house drawing; the blank room information comprises one or more of the size of a room, a door and window, a bay window and the size and the position of an air conditioning hole;
or
And acquiring pictures of the blank room drawings uploaded by a user to obtain the room drawings of the room.
In one embodiment, the RNN sequence model employs a structure in which a plurality of GRU modules are connected in series; the inputting the quantized data into an RNN sequence model for sequence arrangement and outputting an arrangement vector of furniture comprises:
and taking the quantized data as the input of the RNN sequence model, interacting with a plurality of GRU modules in the RNN sequence model, and outputting the arrangement vector of the furniture.
In an embodiment, the method for generating a furniture layout drawing further includes: and training by taking the quantitative data output by the CNN as the input of the RNN to obtain an RNN sequence model.
In one embodiment, the arrangement vector comprises an information code of at least one piece of furniture, and the information code comprises a furniture library ID and the center point coordinate of the piece of furniture on the drawing.
In one embodiment, the searching furniture information of the corresponding furniture from a furniture library storing the decoration drawings according to the arrangement vector, and generating the furniture arrangement drawing of the room according to the furniture information includes: and connecting to the furniture library according to the furniture library ID, inquiring the three-dimensional size of the piece of furniture and the central point coordinate of the piece of furniture on a drawing, determining the central point position of the piece of furniture according to the central point coordinate, and generating a furniture arrangement drawing of the room according to the central point position.
A furniture layout drawing generation apparatus, comprising: the acquisition module is used for acquiring the house drawing of the room;
the extraction module is used for reading the house drawing by using a CNN network and extracting the characteristics to obtain quantitative data of the house; the CNN network is obtained based on collected decoration drawings through training;
the arrangement module is used for inputting the quantized data into an RNN sequence model for sequence arrangement and outputting arrangement vectors of furniture; the RNN sequence model is obtained by training furniture arrangement information based on the decoration drawing;
and the generating module is used for searching the furniture information of the corresponding furniture from a furniture library storing the decoration drawing according to the arrangement vector and generating the furniture arrangement drawing of the room according to the furniture information.
In one embodiment, the device for generating a furniture layout drawing further includes: the system comprises a preprocessing module, a storage module and a display module, wherein the preprocessing module is used for collecting decoration drawings of existing rooms, including house information and furniture information corresponding to the house information; acquiring three-dimensional data and position information of the ground furniture of the decoration drawing; and establishing the furniture library according to the three-dimensional data and the position information thereof.
In one embodiment, the device for generating a furniture layout drawing further includes: the CNN model training module is used for acquiring the collected decoration drawings; deleting the furniture information in the decoration drawing to obtain a picture format drawing containing the house information; and training the CNN network by taking the picture format drawing as input.
In one embodiment, the device for generating a furniture layout drawing further includes: and the RNN sequence model training module is used for training by taking the quantitative data output by the CNN network as the input of the RNN network to obtain the RNN sequence model.
In addition, the application also provides a computer device and a computer readable storage medium.
A computer apparatus, comprising: one or more processors;
a memory;
one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to: the furniture layout drawing generation method is executed.
A computer readable storage medium storing at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by a processor to implement the method for generating a furniture layout drawing.
According to the furniture layout drawing generation method, the furniture layout drawing generation device, the computer equipment and the computer readable storage medium, firstly, a room drawing is obtained, and then a CNN network obtained based on collected decoration drawing training is utilized to perform feature extraction on the room drawing to obtain quantitative data; and inputting the related quantitative data into an RNN sequence model to perform sequence arrangement to obtain an arrangement vector of furniture, and finally searching furniture information of corresponding furniture from a furniture library of the decoration drawing according to the arrangement vector to generate the furniture arrangement drawing of the room. According to the technical scheme, the arrangement scheme of furniture is generated by combining the convolution cyclic neural network, the algorithm speed is higher, and the scheme is more diverse. The CNN network is obtained by training based on the collected decoration drawings, and the drawings of the room do not need to be marked, so that the marking workload is reduced, and the efficiency is improved. Meanwhile, the RNN sequence model is used for carrying out sequence arrangement to realize the generation of arrangement vectors of furniture, the arrangement sequence can refer to factors such as the degree of the requirement of a user on the furniture, the use frequency and the like, for example, the room is considered from an entrance door, and the entrance door is used as an entrance and exit, so that the use effect of the furniture arrangement drawing is better.
In addition, house decoration design drawings of existing professional designers on the market can be collected in actual operation, and a furniture library is built according to the furniture types and sizes on the drawings, so that sample marking and network training work is avoided, and the generated furniture arrangement drawings also avoid the defect that furniture cannot fall to the ground due to stretching.
Additional aspects and advantages of the present application 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 present application.
Drawings
The foregoing and/or additional aspects and advantages of the present application will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a flow chart of a method for generating a furniture layout drawing;
FIG. 2 is a flow chart of another furniture layout drawing generation method;
FIG. 3 is a fitting up drawing of a girl's house training the true value of a sample;
fig. 4 is a schematic structural diagram of a CNN network;
FIG. 5 is a schematic view of information encoding of a piece of furniture;
FIG. 6 is a schematic diagram of the structure of the RNN sequence model;
FIG. 7 is a schematic diagram of a GRU structure;
FIG. 8 is a flowchart of an example application for generating floor furniture layout drawings;
FIG. 9 is a schematic diagram of a furniture layout drawing generation device;
FIG. 10 is a schematic diagram of another furniture layout drawing generation device.
Detailed Description
Reference will now be made in detail to embodiments of the present application, 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 drawings are exemplary only for the purpose of explaining the present application and are not to be construed as limiting the present application.
As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. As used herein, the term "and/or" includes all or any element and all combinations of one or more of the associated listed items.
It will be understood by those within the art that, unless otherwise defined, all terms (including 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. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The generation scheme of furniture layout drawing that this application provided is directed at often need label the drawing in room in traditional furniture layout design generation scheme, and the defect that labeling work load is big, efficiency is lower. The method can directly utilize the ready-made decoration drawing for training, avoids drawing marking work of the traditional scheme, reduces workload and improves efficiency. The house decoration design drawings of professional designers on the market can be collected, and a furniture warehouse is built according to the furniture types and sizes on the drawings, so that not only is unnecessary work avoided, but also the defect that the generated layout drawings cannot fall to the ground to be implemented is overcome.
When implementing this application technical scheme, can operate on a computer equipment, also can operate on the server, perhaps operate on the high in the clouds platform, through deploying on these equipment, the user only need with the house information input in room, can conveniently acquire the furniture of this room and arrange the drawing.
Referring to fig. 1, fig. 1 is a flowchart of a method for generating a furniture layout drawing, including the following steps: and S140, acquiring a house drawing of the room.
In use, the user needs to input the house drawing of the room to finally generate the furniture arrangement drawing, and the arrangement depends on the house information of the room.
In one embodiment, the room may be a blank room, and there may be two main methods for obtaining the house drawing of the room: a. acquiring blank room information input by a user, and converting the blank room information to obtain a house drawing; the blank room information may include room size, windows and doors, bay windows, air conditioning hole size and location, and the like.
b. And directly obtaining the picture of the blank room drawing uploaded by the user to obtain the room drawing.
In practical application, a user can input the information of the blank room, such as the information of the blank room with the size and the position of a room, a door window, a bay window and an air conditioning hole, or directly upload the picture of the drawing of the blank room.
S150, reading the house drawing by using a CNN (Convolutional Neural Network) Network, extracting features, and acquiring quantitative data of the house; the CNN network is obtained by training based on the collected decoration drawings.
For the CNN, the drawing of a room needs to be marked in the traditional scheme, and the marking workload is large. In the embodiment, house decoration drawings designed by professional designers on the market can be fully utilized and collected, a furniture library is established by combining the furniture types and the sizes of the furniture on the decoration drawings, and meanwhile, the collected decoration drawings are utilized to train the CNN network, so that the house drawings in a room do not need to be marked, the workload of marking is reduced, and the generated furniture layout drawings avoid the defect that the furniture cannot fall to the ground for implementation.
S160, inputting the quantized data into an RNN (Recurrent Neural Network) sequence model for sequence arrangement, and outputting an arrangement vector of furniture; the RNN sequence model is obtained by training furniture arrangement information based on the decoration drawing.
The quantitative data generated by the CNN network is input into the RNN sequence model for sequence arrangement, so that an arrangement vector of furniture can be generated, and the arrangement vector is used as an arrangement parameter, so that the generated furniture layout scheme is more in line with the use habit, the use frequency and the like of a user.
Preferably, the arrangement scheme of the furniture can be considered from the entrance door in consideration of the characteristics that the floor furniture is arranged close to the wall and the practicability, and the use habit of a designer is often referred to when designing drawings.
For example, the entrance door of a room can be used as an entrance, the generated floor furniture arrangement scheme is arranged clockwise from the entrance door, and the specific arrangement depends on the information of a blank room; the furniture arrangement is related to the position, size and the like of the entrance door, and the arrangement distance, the arrangement sequence and the like among the furniture also meet the requirement degree, the use frequency and the like of the user on the furniture.
S170, searching furniture information of the corresponding furniture from a furniture library storing the decoration drawing according to the arrangement vector, and generating a furniture arrangement drawing of the room according to the furniture information.
And after the arrangement vector is obtained, connecting the furniture to a furniture library to search the furniture information of the corresponding furniture, wherein the furniture library stores the pre-collected decoration drawing, and finally generating the furniture arrangement drawing of the room by using the searched furniture information.
By combining the technical scheme provided by each embodiment, firstly, house drawings of a room are obtained, and then, the house drawings are subjected to feature extraction by using a CNN network obtained based on collected decoration drawings training to obtain quantitative data; and inputting the related quantitative data into an RNN sequence model to perform sequence arrangement to obtain an arrangement vector of furniture, and finally searching furniture information of corresponding furniture from a furniture library of the decoration drawing according to the arrangement vector to generate the furniture arrangement drawing of the room.
According to the embodiment, the arrangement scheme of the furniture is generated by combining the convolution cyclic neural network, the algorithm speed is higher, and the scheme is more diverse; the CNN network is obtained by training based on the collected decoration drawings, and the drawings of the room do not need to be labeled, so that the labeling workload is reduced, and the efficiency is improved; meanwhile, the RNN sequence model is used for carrying out sequence arrangement to realize the generation of arrangement vectors of furniture, the arrangement sequence can refer to factors such as the degree of the requirement of a user on the furniture, the use frequency and the like, for example, the room is taken as an entrance from an entrance door, so that the use effect of the furniture arrangement drawing is better, and the use habit of the user is better met.
In order to perfect the scheme, the application also provides a plurality of following embodiments; referring to fig. 2, fig. 2 is a flowchart of a method for generating a furniture layout drawing according to another embodiment.
In an embodiment, the method for generating a furniture layout drawing provided by the present application may further include the following steps: s110, collecting decoration drawings of existing rooms, wherein the decoration drawings comprise house information and furniture information corresponding to the house information; acquiring three-dimensional data and position information of the ground furniture of the decoration drawing; and establishing the furniture library according to the three-dimensional data and the position information thereof.
Specifically, the embodiment further includes a training data preprocessing process, which collects various room decoration schemes, obtains three-dimensional sizes of room floor furniture, and creates a furniture library, as shown in table 1 below: TABLE 1
Further, this embodiment may further include:
and S120, acquiring the collected decoration drawing, deleting the furniture information in the decoration drawing to obtain a picture format drawing containing the house information, and training the CNN network by taking the picture format drawing as input.
Specifically, a CNN network is trained, collected decoration drawings of a room are utilized, the decoration drawings comprise house information and corresponding furniture information, and the decoration drawings are used as true values for training the CNN network; and after the furniture information is deleted, obtaining a picture format drawing only containing house information (such as the sizes and positions of rooms, doors and windows, bay windows and air conditioning holes) and taking the drawing as a drawing sample to train the CNN network.
For the training of the training drawing sample, referring to fig. 3, fig. 3 is a fitting-up drawing of the girl's house with the real value of the training sample, and the right drawing is a blank house drawing of the girl's house containing only blank house information such as rooms, doors and windows, bay windows and the like, and is used as the drawing sample.
As an embodiment, the scheme for training the CNN network in S120 above may further include: inputting the picture format drawing into a CNN network; and performing feature extraction on the picture format drawing through a plurality of convolution cores of the CNN network, and converting the extracted house information into an intermediate state through nonlinear transformation.
Because manual labeling workload is large and mistakes are easy to make, by using the blank room drawing in the picture format as the drawing sample of the CNN network, the blank room drawing in the picture format can be used as input during training, and when the blank room drawing in the picture format is used, a user uploads blank room information and converts the blank room information into the blank room drawing in the picture format, or the user directly uploads the blank room drawing in the picture format.
As for the structure of the CNN network, referring to fig. 4, fig. 4 is a schematic diagram of the structure of the CNN network; the structural schematic diagram describes the process of CNN network extraction characteristics and the process of generating arrangement vectors for RNN sequence models, after the blank house picture paper in picture format is transmitted into the CNN network, the drawing paper is extracted by a plurality of turns of convolution layers and pooling layers, and is spread in rows to be connected into vectors and transmitted into a full connection layer, drawing information characteristics are extracted, the characteristics of the blank house are completely stored in a vector form, and the intermediate state is used as the input of the RNN network sequence models.
Through the scheme of the embodiment, the effect of inputting by using the picture is better in blank room information processing, the blank room drawing in the picture format is more direct than the blank room description information, the CNN network performs convolution on the whole drawing, the convolution kernel reduces parameters, and the training speed is higher under the conditions of huge data volume and high real-time requirement.
In one embodiment, the RNN sequence model employs a structure in which a plurality of GRU modules are connected in series; in the RNN sequence model training, the method may further include: and S130, training by taking the quantized data output by the CNN as the input of the RNN to obtain an RNN sequence model.
Correspondingly, when the quantized data is input into the RNN sequence model for sequence arrangement and output of the arrangement vector of the room in S160, the method may include:
and taking the quantized data as the input of an RNN sequence model, interacting through a plurality of GRU modules in the RNN sequence model, and outputting the arrangement vector of the furniture.
Further, the step of searching the furniture information of the corresponding furniture from the furniture library storing the decoration drawing according to the arrangement vector in S170, and generating the furniture arrangement drawing of the room according to the furniture information may include: and connecting to the furniture library according to the furniture library ID, inquiring the three-dimensional size of the piece of furniture and the central point coordinate of the piece of furniture on a drawing, determining the central point position of the piece of furniture according to the central point coordinate, and generating a furniture arrangement drawing of the room according to the central point position.
In one embodiment, the arrangement vector may include an information code of at least one piece of furniture, the information code including a furniture library ID and coordinates of a center point of the piece of furniture on a drawing sheet. In this embodiment, the CNN network is an encoder that encodes the house characteristic information on the house drawing into an intermediate state value. The RNN sequence model is a decoder that predicts furniture arrangement sequences based on house characteristic information.
The ground furniture sequences s1, s2, … sI' thus output are based on the intermediate state and the output of the last GRU module.
The error of the CNN network is derived from the cross entropy of the predicted value of the network and the actual value of the drawing, namely:
wherein p (xi) is the distribution position of the floor furniture arrangement on the real drawing, and q (si) is the distribution position of the furniture arrangement predicted by the RNN network.
According to the scheme of the embodiment, quantitative data of the blank room are obtained through characteristic information of the CNN, and then the quantitative data are used as input to predict the floor furniture layout drawing of the room through strong nonlinearity and memory capacity of the RNN.
As an example, in the arrangement vector code, the information code of a piece of furniture is 12 bits, the front 4 bits are the ID of the furniture library, and the rear 8 bits are the millimeter coordinates of the center point of the piece of furniture on the drawing. The ID codes of the furniture in the furniture warehouse are unified into 4 digits, the first two digits are codes of the furniture storage space, and the last two digits are furniture category codes; as shown in fig. 5, fig. 5 is a schematic diagram of information coding of a piece of furniture, 03 represents a living room, 09 represents a tea table, the three-dimensional size of the tea table in the living room can be queried through 0309, and the three-dimensional size information of the tea table in the living room can be queried through 0309; 15382170 shows the center point (x, y) of the table on the drawing as (1538, 2170) mm.
For the structure of the RNN sequence model, as shown in fig. 6, fig. 6 is a schematic diagram of the RNN sequence model, and the initial input of the model is the intermediate state of the CNN network output; after the multi-round hidden state processing, the output Si of the last GRU module is used as the input of the current GRU, the input value of the _DOORentry gate vector is used as the starting of RNN sequence model arrangement, and the _EOSis finally output as a furniture arrangement result.
As for the structure of the GRU module, referring to fig. 7, fig. 7 is a schematic structural diagram of the GRU, and in the present application, information is transmitted directly using a hidden state through the GRU module in an RNN network, and forgetting and selecting information are performed simultaneously by using the same gate control (update gate).
In order to make the technical solution of the present application clearer, an application example of a floor furniture layout drawing of the present application is described below by taking a blank room as an example, and referring to fig. 8, fig. 8 is a flowchart of an application example of a floor furniture layout drawing, and includes:
s801, acquiring a house drawing picture of the blank house input by the user, and turning to s 804;
s802, acquiring blank room information input by a user;
s803, converting the blank room information into a drawing;
s804, generating a blank room drawing, and turning to s 806;
s805, collecting decoration drawings of the room, turning to s806, and executing s 810;
s806, inputting the CNN network to extract image features;
s807, encoding the feature information into an intermediate state value state;
s808, inputting an RNN sequence model for arrangement;
s809, generating the arrangement vector of the floor furniture, and turning to s 812;
s810, acquiring three-dimensional data of the floor furniture;
s811, create furniture store, go to s 812;
s812, inquiring the furniture information of the floor furniture according to the arrangement vector;
s813, determining the position of the center point of the floor furniture according to the coordinates of the arrangement vector;
and s814, generating and outputting the floor furniture arrangement drawing of the blank room.
Through the above example, the user can input the blank room information, and also can upload the blank room drawing, namely the furniture arrangement drawing of the blank room.
Referring to fig. 9, fig. 9 is a schematic structural diagram of a device for generating a furniture layout drawing, which mainly includes: the acquisition module 140 is used for acquiring a house drawing of a room;
the extraction module 150 is configured to read the house drawing by using a CNN network, perform feature extraction, and acquire quantized data of the house; the CNN network is obtained based on collected decoration drawings through training;
the arrangement module 160 is configured to input the quantized data into an RNN sequence model for sequence arrangement, and output an arrangement vector of furniture; the RNN sequence model is obtained by training furniture arrangement information based on the decoration drawing;
and the generating module 170 is configured to search the furniture information of the corresponding furniture from a furniture library storing the decoration drawing according to the arrangement vector, and generate the furniture arrangement drawing of the room according to the furniture information.
Referring to fig. 10, fig. 10 is a schematic structural diagram of another furniture layout drawing generation device, further including: in one embodiment, the device for generating a furniture layout drawing further includes: the preprocessing module 110 is configured to collect decoration drawings of existing rooms, which include house information and furniture information corresponding to the house information; acquiring three-dimensional data and position information of the ground furniture of the decoration drawing; and establishing the furniture library according to the three-dimensional data and the position information thereof.
In one embodiment, the device for generating a furniture layout drawing further includes: a CNN model training module 120, configured to obtain the collected decoration drawing; deleting the furniture information in the decoration drawing to obtain a picture format drawing containing the house information; and training the CNN network by taking the picture format drawing as input.
In one embodiment, the device for generating a furniture layout drawing further includes: and an RNN sequence model training module 130, configured to train with the quantized data output by the CNN network as an input of the RNN network to obtain an RNN sequence model.
A computer apparatus, comprising: one or more processors;
a memory;
one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to: the furniture layout drawing generation method is executed.
A computer readable storage medium storing at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by a processor to implement the method for generating a furniture layout drawing.
By combining the above embodiments, the arrangement scheme of the floor furniture can be directly generated more quickly, and the scheme has diversity; a user only needs to provide blank room information in the whole process, and other operations are not needed, so that the design drawing meeting the basic use requirement can be obtained. The training time of the RNN sequence model is saved by adopting the GRU structure, and the training efficiency is improved. The floor furniture is considered from the practicality in arranging, starts considering from entering the door to go into the door and regard as the room access & exit, and during the drawing design, the furniture is arranged and is gone into the position of door, size etc. and is relevant, and arranges interval, the order of arranging etc. and also accords with user's demand degree, frequency of use etc. to furniture between the furniture.
It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and may be performed in other orders unless explicitly stated herein. Moreover, at least a portion of the steps in the flow chart of the figure may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
The foregoing is only a partial embodiment of the present application, and it should be noted that, for those skilled in the art, several modifications and decorations can be made without departing from the principle of the present application, and these modifications and decorations should also be regarded as the protection scope of the present application.
Claims (13)
1. A method for generating a furniture layout drawing is characterized by comprising the following steps:
acquiring a house drawing of a room;
reading the house drawing by using a CNN network, extracting features, and acquiring quantitative data of the house; the CNN network is obtained based on collected decoration drawings through training;
inputting the quantized data into an RNN sequence model for sequence arrangement, and outputting arrangement vectors of furniture; the RNN sequence model is obtained by training furniture arrangement information based on the decoration drawing;
searching furniture information of corresponding furniture from a furniture library storing the decoration drawing according to the arrangement vector, and generating a furniture arrangement drawing of the room according to the furniture information;
the arrangement vector comprises an information code of at least one piece of furniture, and the information code comprises a furniture library ID of the piece of furniture in the furniture library and a center point coordinate of the piece of furniture on a drawing;
the searching the furniture information of the corresponding furniture from the furniture library storing the decoration drawing according to the arrangement vector and generating the furniture arrangement drawing of the room according to the furniture information comprises the following steps:
connecting to the furniture library according to the furniture library ID, inquiring the three-dimensional size of the furniture and the central point coordinate of the furniture on a drawing, determining the central point position of the furniture according to the central point coordinate, and generating a furniture arrangement drawing of the room according to the central point position;
wherein the RNN sequence model is a recurrent neural network sequence model.
2. The method for generating a furniture layout drawing according to claim 1, further comprising:
collecting decoration drawings of existing rooms, wherein the decoration drawings comprise house information and furniture information corresponding to the house information;
acquiring three-dimensional data and position information of the ground furniture of the decoration drawing;
and establishing the furniture library according to the three-dimensional data and the position information thereof.
3. The method for generating a furniture layout drawing according to claim 2, further comprising:
acquiring the collected decoration drawings;
deleting the furniture information in the decoration drawing to obtain a picture format drawing containing the house information;
and training the CNN network by taking the picture format drawing as input.
4. The method for generating a furniture layout drawing of claim 3, wherein the training of the CNN network with the picture format drawing as an input comprises:
inputting the picture format drawing into a CNN network;
and performing feature extraction on the picture format drawing through a plurality of convolution cores of the CNN network, and converting the extracted house information into an intermediate state through nonlinear transformation.
5. The method for generating the furniture layout drawing according to claim 1, wherein the room is a blank room;
the acquiring of the house drawing of the room comprises the following steps:
acquiring blank room information input by a user, and converting the blank room information to obtain a house drawing; the blank room information comprises one or more of the size of a room, a door and window, a bay window and the size and the position of an air conditioning hole;
or
And acquiring pictures of the blank room drawings uploaded by a user to obtain the room drawings of the room.
6. The method for generating the furniture layout drawing of claim 4, wherein the RNN sequence model adopts a structure in which a plurality of GRU modules are connected in series;
the inputting the quantized data into an RNN sequence model for sequence arrangement and outputting an arrangement vector of furniture comprises:
and taking the quantized data as the input of the RNN sequence model, interacting with a plurality of GRU modules in the RNN sequence model, and outputting the arrangement vector of the furniture.
7. The method for generating a furniture layout drawing according to claim 6, further comprising:
and training by taking the quantitative data output by the CNN as the input of the RNN to obtain an RNN sequence model.
8. A furniture layout drawing generation device is characterized by comprising:
the acquisition module is used for acquiring the house drawing of the room;
the extraction module is used for reading the house drawing by using a CNN network and extracting the characteristics to obtain quantitative data of the house; the CNN network is obtained based on collected decoration drawings through training;
the arrangement module is used for inputting the quantized data into an RNN sequence model for sequence arrangement and outputting arrangement vectors of furniture; the RNN sequence model is obtained by training furniture arrangement information based on the decoration drawing;
the generating module is used for searching furniture information of corresponding furniture from a furniture library storing the decoration drawing according to the arrangement vector and generating a furniture arrangement drawing of the room according to the furniture information;
the arrangement vector comprises an information code of at least one piece of furniture, and the information code comprises a furniture library ID of the piece of furniture in the furniture library and a center point coordinate of the piece of furniture on a drawing;
the generation module is further used for connecting to the furniture library according to the furniture library ID, inquiring the three-dimensional size of the piece of furniture and the central point coordinate of the piece of furniture on a drawing, determining the central point position of the piece of furniture according to the central point coordinate, and generating a furniture arrangement drawing of the room according to the central point position;
wherein the RNN sequence model is a recurrent neural network sequence model.
9. The apparatus for generating furniture layout drawing of claim 8, further comprising:
the system comprises a preprocessing module, a storage module and a display module, wherein the preprocessing module is used for collecting decoration drawings of existing rooms, including house information and furniture information corresponding to the house information; acquiring three-dimensional data and position information of the ground furniture of the decoration drawing; and establishing the furniture library according to the three-dimensional data and the position information thereof.
10. The apparatus for generating a furniture layout drawing of claim 9, further comprising:
the CNN model training module is used for acquiring the collected decoration drawings; deleting the furniture information in the decoration drawing to obtain a picture format drawing containing the house information; and training the CNN network by taking the picture format drawing as input.
11. The apparatus for generating a furniture layout drawing of claim 10, further comprising:
and the RNN sequence model training module is used for training by taking the quantitative data output by the CNN network as the input of the RNN network to obtain the RNN sequence model.
12. A computer device, comprising:
one or more processors;
a memory;
one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to: executing the method of generating a furniture layout according to any of claims 1 to 7.
13. A computer-readable storage medium, characterized in that it stores at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by a processor to implement the generation method of a furniture layout drawing according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911345314.4A CN110751721B (en) | 2019-12-24 | 2019-12-24 | Furniture layout drawing generation method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911345314.4A CN110751721B (en) | 2019-12-24 | 2019-12-24 | Furniture layout drawing generation method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110751721A CN110751721A (en) | 2020-02-04 |
| CN110751721B true CN110751721B (en) | 2020-10-30 |
Family
ID=69285991
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911345314.4A Active CN110751721B (en) | 2019-12-24 | 2019-12-24 | Furniture layout drawing generation method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110751721B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111553012A (en) * | 2020-04-28 | 2020-08-18 | 广东博智林机器人有限公司 | Home decoration design method and device, electronic equipment and storage medium |
| CN111680421A (en) * | 2020-06-05 | 2020-09-18 | 广东博智林机器人有限公司 | Home decoration design method and device, electronic equipment and storage medium |
| CN111882644A (en) * | 2020-06-23 | 2020-11-03 | 北京城市网邻信息技术有限公司 | Furniture display method and device |
| CN112257272B (en) * | 2020-10-26 | 2024-01-23 | 南京国豪家装饰设计有限公司 | Decoration design method of furniture set |
| CN113282999B (en) * | 2021-06-25 | 2023-12-22 | 广东都市建筑规划设计有限公司 | Automatic modification generation method and device for electrical lighting drawing and computer equipment |
| CN113343593B (en) * | 2021-08-05 | 2021-10-29 | 佛山市陶风互联网络科技有限公司 | Home decoration wiring planning management method, system and storage medium based on neural network |
| CN113792358A (en) * | 2021-09-22 | 2021-12-14 | 深圳须弥云图空间科技有限公司 | Automatic interaction layout method and device for three-dimensional furniture and electronic equipment |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3596659A4 (en) * | 2017-03-17 | 2021-01-27 | Magic Leap, Inc. | Room layout estimation methods and techniques |
| CN108984904B (en) * | 2018-07-17 | 2022-09-20 | 北京理工大学 | Home design method based on deep neural network |
| CN109272123B (en) * | 2018-08-03 | 2021-06-22 | 常州大学 | Sucker-rod pump working condition early warning method based on convolution-circulation neural network |
| CN109344815B (en) * | 2018-12-13 | 2021-08-13 | 深源恒际科技有限公司 | Document image classification method |
| CN109740243B (en) * | 2018-12-29 | 2022-07-08 | 江苏艾佳家居用品有限公司 | Furniture layout method and system based on piece-by-piece reinforcement learning technology |
| CN109871604B (en) * | 2019-01-31 | 2023-05-16 | 浙江工商大学 | Indoor function division method based on depth countermeasure network model |
| CN110046226B (en) * | 2019-04-17 | 2021-09-24 | 桂林电子科技大学 | Image description method based on distributed word vector CNN-RNN network |
| CN110363853B (en) * | 2019-07-15 | 2020-09-01 | 贝壳找房(北京)科技有限公司 | Furniture placement scheme generation method, device and equipment and storage medium |
| KR20190106867A (en) * | 2019-08-27 | 2019-09-18 | 엘지전자 주식회사 | An artificial intelligence apparatus for guiding arrangement location of furniture and operating method thereof |
| CN110569842B (en) * | 2019-09-05 | 2022-08-12 | 江苏艾佳家居用品有限公司 | Semi-supervised learning method for GAN model training |
-
2019
- 2019-12-24 CN CN201911345314.4A patent/CN110751721B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110751721A (en) | 2020-02-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110751721B (en) | Furniture layout drawing generation method and device, computer equipment and storage medium | |
| Dai et al. | Scannet: Richly-annotated 3d reconstructions of indoor scenes | |
| CN107229757B (en) | Video retrieval method based on deep learning and Hash coding | |
| Wu et al. | Visual place categorization: Problem, dataset, and algorithm | |
| EP3867883A1 (en) | 3d scene synthesis techniques using neural network architectures | |
| CN112037312B (en) | Real-time human body posture inverse kinematics solving method and device | |
| Ehrhardt et al. | Learning a physical long-term predictor | |
| CN112016601B (en) | Network model construction method based on knowledge graph enhanced small sample visual classification | |
| CN112132197A (en) | Model training method, image processing method, device, computer equipment and storage medium | |
| CN115223067B (en) | Point cloud fusion method, device and equipment applied to unmanned aerial vehicle and storage medium | |
| CN116402671B (en) | Sample coding image processing method for automatic coding system | |
| CN112668492A (en) | Behavior identification method for self-supervised learning and skeletal information | |
| CN113051420A (en) | Robot vision man-machine interaction method and system based on text generation video | |
| Liu et al. | Atvio: Attention guided visual-inertial odometry | |
| CN110210523B (en) | Method and device for generating image of clothes worn by model based on shape graph constraint | |
| Kotar et al. | Entl: Embodied navigation trajectory learner | |
| Reza et al. | Automatic annotation for semantic segmentation in indoor scenes | |
| CN115482557B (en) | Human body image generation method, system, equipment and storage medium | |
| CN110298912A (en) | Reproducing method, system, electronic device and the storage medium of three-dimensional scenic | |
| Ehrhardt et al. | Learning to represent mechanics via long-term extrapolation and interpolation | |
| Yan et al. | An interactive system for efficient 3D furniture arrangement | |
| CN110738640B (en) | Spatial data comparison method and related product | |
| Majumdar et al. | Ssl enables learning from sparse rewards in image-goal navigation | |
| CN106503174A (en) | A kind of environment Visualization method and system modeled based on Network Three-dimensional | |
| Dong et al. | PanoContext-Former: Panoramic Total Scene Understanding with a Transformer |
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 |