Building design method, device and equipment based on deep learning and medium

Abstract

The invention discloses a building design method and device based on deep learning, equipment and a medium, and the method comprises the steps: obtaining an initial training model and to-be-trained building model data, and the initial training model comprises at least one design condition; inputting the to-be-trained building model data into the at least one design condition of the initial training model for training to obtain a target training model; obtaining at least one design parameter of a target building, wherein each design parameter is generated after a user assigns at least one design condition of the target building; and inputting the at least one design parameter into the target training model for training, and generating a target building model corresponding to the target building after the design of the at least one design condition, so that the building model meeting the condition can be quickly and accurately generated through the model training method, the design efficiency and precision are improved, and the design cost is reduced. And meanwhile, a greater design freedom degree is provided for a user designer.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering modeling, and particularly relates to an architectural design method, device, equipment and medium based on deep learning. Background Art

[0002] The architectural design process is a very complex process. The current architectural design methods are divided into three categories: manual design, parametric design, and generative design. Manual design is to manually complete the whole design process with a pen, paper or design software.

[0003] Parametric design is to express the relationship between building components through a structure diagram, so that when one component is changed after the design is completed, other related components can change together. Generative design is to artificially design some expected goals, and then through some optimization algorithms, such as genetic algorithms with random properties, to continuously optimize the current design to achieve some better design results. Its basic principle is to use algorithms and data to automatically or semi-automatically generate design schemes of items or structures. These algorithms and data can be based on different rules and parameters to achieve different design goals and constraints. For example, generative design can be used to automatically generate houses or sites with specific aesthetic features, or to optimize items or structures with a specific function or performance. However, the existing design methods are inefficient and cannot meet the user's needs. Summary of the Invention

[0004] In view of this, the present invention provides an architectural design method based on deep learning to solve the problem of low efficiency existing in the existing design methods.

[0005] In a first aspect, the present invention provides an architectural design method based on deep learning, the method comprising: obtaining an initial training model and building model data to be trained, the initial training model including at least one design condition; inputting the building model data to be trained into the at least one design condition of the initial training model for training to obtain a target training model; obtaining at least one design parameter of a target building, each design parameter being generated after a user assigns values to at least one design condition of the target building; inputting the at least one design parameter into the target training model for training, and after being designed by the at least one design condition, generating a target building model corresponding to the target building.

[0006] In this embodiment, first, an initial training model and building model data are obtained, and these data are input into the initial model for training to obtain a target training model. Then, design parameters of the target building are obtained, and these parameters are assigned by the user according to the design conditions. Next, these design parameters are input into the target training model for training to generate a building model corresponding to the target building. This method of using a training model can quickly and accurately generate a building model that meets the conditions, improving the efficiency and accuracy of the design. At the same time, it also provides greater design freedom for user designers.

[0007] Combined with the first aspect, in an optional implementation manner, after inputting each of the design parameters into the first training model for training to generate the building model corresponding to the target building, the method further includes: receiving a modification operation by the user on some model content in the building model; in response to the modification operation by the user, modifying the design parameters corresponding to the building model to generate at least one first design parameter; and adaptively modifying the remaining unchanged part of the building model according to the at least one first design parameter to generate a second building model of the target building.

[0008] In this embodiment, after generating the building model, by receiving the modification operation of the user on the building model and automatically adjusting the remaining unchanged part to adapt to these modifications, a second building model is generated. This method not only considers the specific needs of the user but also makes adaptive adjustments according to the modification operation of the user to ensure the consistency and coordination of the design. This provides greater design freedom and personalized customization ability for the user while maintaining the integrity and efficiency of the design.

[0009] Combined with the first aspect, in an optional implementation manner, the adaptively modifying the remaining unchanged part of the building model according to the at least one first design parameter to generate the second building model of the target building includes: obtaining at least one second design parameter of the building model, where the at least one second design parameter is the parameter corresponding to the remaining unchanged part of the building model; sorting the second design parameters to generate a second design sequence; and modifying the building model in the order of the second design sequence to generate the second building model of the target building.

[0010] In this embodiment, by obtaining the second design parameters of the building model and sorting them according to a certain rule to generate a second design sequence. Then, modifying the building model according to the order of the second design sequence to generate the second building model of the target building. This method takes into account the specific needs of the user and can adaptively modify the remaining unchanged part to maintain the consistency and coordination of the design. Through the steps of sorting and modifying one by one, the design process can be adjusted and optimized more flexibly to meet different design requirements and conditions.

[0011] In combination with the first aspect, in an alternative embodiment, after obtaining at least one design parameter of the target building, where each design parameter is generated after a user assigns values to at least one design condition of the target building, the method further includes: arranging each design parameter in a preset order to generate at least one design sequence.

[0012] In combination with the first aspect, in an alternative embodiment, before obtaining the initial training model, the method further includes: obtaining at least one design condition of the target building; generating at least one design sequence according to the at least one design condition, where each sequence corresponds to one or more design conditions; and generating the initial training model according to the at least one sequence.

[0013] In this embodiment, before obtaining the initial training model, it is necessary to first obtain at least one design condition of the target building. According to these design conditions, at least one design sequence can be generated, and each sequence corresponds to one or more design conditions. Finally, based on these sequences, the initial training model can be generated. This embodiment reflects the importance of considering various different building elements in the design process and the necessity of integrating these elements into the model.

[0014] In combination with the first aspect, in an alternative embodiment, the method further includes: before the training, the initial training model includes the at least one design condition, the at least one design condition corresponds to a fixed value, and the user is allowed to input a fixed numerical value in the at least one design condition; after the training, the target training model includes the at least one design condition, the at least one design condition corresponds to one or more values, and the user is allowed to input any numerical value in the at least one design condition.

[0015] In this embodiment, before the training, the initial training model includes at least one design condition, and the user can input a fixed positive numerical value for optimization. After the training, the target training model includes at least one design condition, allowing the user to input any numerical value. This method improves the adaptability and flexibility of the model, making it better meet the actual needs.

[0016] In combination with the first aspect, in an alternative embodiment, after the training, the target training model includes at least one design condition, and the method further includes: after the training, the target training model includes at least one design condition including a regression loss, and the regression loss is a result calculated based on the at least one design parameter.

[0017] In this embodiment, the trained target training model includes at least one design condition, which is associated with the regression loss. The regression loss is a result calculated based on at least one design parameter and is used to evaluate the performance of the model. In this way, the performance of the model under different design conditions can be better understood, and corresponding adjustments and optimizations can be made to improve the prediction and decision-making capabilities of the model.

[0018] In a second aspect, the present invention provides a building design device based on deep learning, the device comprising:

[0019] A first acquisition module, configured to acquire an initial training model and building model data to be trained, the initial training model including at least one design condition;

[0020] A training module, configured to input the building model data to be trained into the at least one design condition of the initial training model for training to obtain a target training model;

[0021] A second acquisition module, configured to acquire at least one design parameter of a target building, each design parameter being generated after a user assigns values to at least one design condition of the target building;

[0022] A generation module, configured to input the at least one design parameter into the target training model for training, and generate a target building model corresponding to the target building after designing through the at least one design condition.

[0023] In a third aspect, the present invention provides an electronic device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to execute a building design method based on deep learning according to the first aspect or any corresponding embodiment thereof.

[0024] In a fourth aspect, the present invention provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to cause a computer to execute a building design method based on deep learning according to the first aspect or any corresponding embodiment thereof. Description of the Drawings

[0025] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0026] Figure 1(a) is a flowchart of a building design method based on deep learning according to the present invention;

[0027] Figure 1(b) is a schematic diagram of a building design method based on deep learning according to the present invention;

[0028] Figure 2(a) is a flowchart of another building design method based on deep learning according to the present invention;

[0029] Figure 2(b) is a flowchart of yet another building design method based on deep learning according to the present invention;

[0030] Figure 3(a) is a schematic diagram of another building design method based on deep learning according to the present invention;

[0031] Figure 3(b) is a schematic diagram of yet another building design method based on deep learning according to the present invention;

[0032] Figure 3(c) is a schematic diagram of yet another building design method based on deep learning according to the present invention;

[0033] Figure 4(a) is a flowchart of yet another building design method based on deep learning according to the present invention;

[0034] Figure 4(b) is a schematic diagram of yet another building design method based on deep learning according to the present invention;

[0035] Figure 4(c) is a schematic diagram of yet another building design method based on deep learning according to the present invention;

[0036] Figure 5(a) is a schematic diagram of yet another building design method based on deep learning according to the present invention;

[0037] Figure 5(b) is a schematic diagram of yet another building design method based on deep learning according to the present invention;

[0038] Figure 5(c) is a schematic diagram of yet another building design method based on deep learning according to the present invention;

[0039] Figure 6 is a structural block diagram of a building design device based on deep learning according to the present invention;

[0040] Figure 7 is a schematic diagram of the hardware structure of the electronic device of the present invention. Detailed implementation manners

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0042] An embodiment of the present invention provides a building design method based on deep learning, which is used to solve the problem of low efficiency existing in the existing design methods.

[0043] According to an embodiment of the present invention, an embodiment of a building design method based on deep learning is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0044] In this embodiment, a component statistics method based on a building model is provided, which can be used for the above-mentioned mobile terminals, such as PCs, tablets, etc. Fig. 1(a) is a flowchart of a building design method based on deep learning according to an embodiment of the present invention. As shown in Fig. 1(a), the process includes the following steps:

[0045] Step 101, obtain an initial training model and building model data to be trained, where the initial training model includes at least one design condition.

[0046] Specifically, obtaining an initial training model and building model data to be trained means modifying a character embedding model based on a transformer into a building entity model and adding corresponding constraint conditions to the model. The constraint conditions refer to at least one design condition, such as plot length, floor area ratio, and total number of buildings, etc. The data to be trained refers to some existing design cases, and the design cases are building models that have completed the design.

[0047] The Transformer model is a classic model in NLP (Natural Language Processing). Its characteristic is that it abandons the traditional CNN (Convolutional Neural Networks) and RNN (Recurrent Neural Network). The entire network structure is completely composed of attention mechanisms. Its advantages are that it can perform parallel computing with high computational efficiency and can also consider the global information of the data. However, at the same time, the Transformer model also has some disadvantages, such as a large amount of computation, many training parameters, a small receptive field, and being affected by the window size of the convolutional kernel, and it cannot consider the entire context simultaneously. But for building entities, there is no semantic set like the Transformer. For example, for the building concept of plot length, theoretically, it can take all values from 0 to infinity, so it is impossible to create a dictionary similar to Chinese characters or letters in the Transformer model.

[0048] Step 102: Input the building model data to be trained into at least one of the design conditions of the initial training model for training to obtain a target training model.

[0049] Specifically, input the model data to be trained into the initial model for training to obtain a target training model. The target training model refers to a model that can convert design conditions into a building model. As shown in Figure 1(b), the drawing generation model is the target training model.

[0050] Step 103: Obtain at least one design parameter of the target building, and each design parameter is generated after the user assigns values to at least one design condition of the target building.

[0051] Specifically, the user needs to input the parameters corresponding to the design conditions. For example, the plot length is 10 meters, the total number of buildings is 5, etc., and assign values to the design parameters that the user needs to assign. As shown in Figure 1(b), the constraint conditions of the drawing are the design conditions, such as plot size, floor area ratio, and the number of buildings, etc.

[0052] After obtaining at least one design parameter of the target building, and each design parameter is generated after the user assigns values to at least one design condition of the target building, it further includes: arranging each design parameter in a preset order to generate at least one design sequence. It means rearranging the design parameters assigned by the user, arranging the design parameters with values in the front, and arranging the unassigned design conditions at the back.

[0053] Step 104: Input the at least one design parameter into the target training model for training. After being designed by the at least one design condition, a target building model corresponding to the target building is generated.

[0054] Specifically, input the design parameters assigned by the user into the target training model for training. After being assigned by the at least one design condition, the model will generate a target building model corresponding to the target building according to the design conditions input by the user and the parameters corresponding to the design conditions. The target building model can be a three-dimensional BIM (Building Information Modeling) model or a two-dimensional CAD (Computer Aided Drafting) model.

[0055] In this embodiment, first, obtain the initial training model and the data of the building model to be trained, and input these data into the initial model for training to obtain the target training model. Then, obtain the design parameters of the target building, which are assigned by the user according to the design conditions. Then, input these design parameters into the first training model for training to generate a building model corresponding to the target building. This method of using a training model can quickly and accurately generate a building model that meets the conditions and improve the design efficiency.

[0056] In this embodiment, a building design method based on deep learning is provided, which can be used for the above-mentioned mobile terminals, such as mobile phones, tablets, etc. Fig. 2(a) is a flowchart of the building design method based on deep learning according to an embodiment of the present invention. As shown in Fig. 2(a), the process includes the following steps:

[0057] Step 201: Receive a modification operation of the user on a part of the model content in the building model.

[0058] Specifically, when the user is not satisfied with a part of the model content in the building model generated by the target model, the model can be dragged and modified using a mouse. For example, as shown in Fig. 3(a), Fig. 3(a) is the building model generated by the target model, and Fig. 3(b) is the corresponding modification operation when the user is not satisfied with the living room, which is a modification operation on a part of the model content in the building model.

[0059] Step 202: In response to the modification operation of the user, modify the design parameters corresponding to the building model to generate at least one first design parameter.

[0060] Specifically, after the system internally performs a dragging operation on the living room area shown in Figure 3(a) of the user, it will modify the corresponding design parameters according to the pre-input design conditions, such as three bedrooms and one living room, the length of the living room, the width of the living room, etc. For example, in Figure 3(b), the width of the living room is modified from 6 meters to 7 meters, and finally an image shown in Figure 3(c) is generated. These modified parameters are the design parameters corresponding to the user's dragging and modification. The modification can also be to fix the model.

[0061] Step 203, according to the at least one first design parameter, adaptively modify the remaining unchanged parts of the building model to generate a second building model of the target building.

[0062] Specifically, according to the first design parameter, adaptively modify the remaining unchanged parts to generate a second building model of the new target building. Here, the system receives the user's generation instruction, and according to the pre-input constraint conditions, such as 90 square meters, three bedrooms and one living room, etc., and the part of the generation result fixed by the user, such as fixing the bathroom, kitchen, and the adjusted living room, as shown in Figure 3(c), regenerate the remaining parts, namely Bedroom 1 and Bedroom 2. This process can be repeatedly executed until a building model satisfactory to the customer appears.

[0063] In this embodiment, after generating the building model, by receiving the user's modification operation on the building model and automatically adjusting the remaining unchanged parts to adapt to these modifications, a second building model is generated. This method not only considers the specific needs of the user, but also makes adaptive adjustments according to the user's modification operation to ensure the consistency and coordination of the design. This provides the user with greater design freedom and personalized customization ability, while maintaining the integrity and efficiency of the design.

[0064] Specifically, as shown in Figure 2(b), the above step 203 includes:

[0065] Step 2031, obtain at least one second design parameter of the building model, where the at least one second design parameter is a parameter corresponding to the remaining unchanged parts of the building model.

[0066] Specifically, the system needs to obtain the design parameters corresponding to the model not modified by the user, that is, excluding the design parameters corresponding to the model modified by the user and the design parameters corresponding to the model fixed by the user, to generate the second design parameter. For example, as shown in Figure 3(b), the customer modifies the living room, then fixes the bathroom, kitchen, and the adjusted living room, and the design parameters corresponding to the remaining parts, namely Bedroom 1 and Bedroom 2, are the second design parameters.

[0067] Step 2032, sort the second design parameters to generate a second design sequence.

[0068] Specifically, arrange the second design parameters to generate a new sequence, which is the second design sequence. For example, Bedroom 1, Bedroom 2, etc. are arranged in order to form a sequence.

[0069] Step 2033, modify the building model according to the order of the second design sequence to generate the second building model of the target building.

[0070] Specifically, modify the building model according to the order of the second design sequence to generate the second building model of the new target building. Importantly, all building entities, which are rooms in this example, are generated one by one in order. For example, if the generation order is bathroom, living room, kitchen, Bedroom 1, Bedroom 2, then by modifying and fixing the previous part of the generation sequence, the subsequent part can be generated one by one again.

[0071] In this embodiment, by obtaining the second design parameters of the building model, the second design sequence is generated according to certain rules. Then, modify the building model according to the order of the second design sequence to generate the second building model of the target building. This method takes into account the specific needs of the user, can adaptively modify the remaining unmodified parts, and maintain the consistency and coordination of the design. Through the steps of sorting and modifying one by one, the design process can be adjusted and optimized more flexibly to meet different design requirements and conditions.

[0072] In some alternative embodiments, as shown in FIG. 4(a), the above step 101 includes:

[0073] Step 401, obtain at least one design condition of the target building.

[0074] Specifically, obtaining the design condition of the target building is the constraint condition. For example, plot size, floor area ratio, or number of buildings, etc.

[0075] Step 402, generate at least one design sequence according to the at least one design condition, and each sequence corresponds to one or more design conditions.

[0076] Specifically, as shown in FIG. 4(b), generate a design sequence according to each design condition. A design sequence can have one or more conditions. For example, plot length, plot width, floor area ratio, number of buildings, bedroom length, and living room length, etc.

[0077] Step 403, generate the initial training model according to the at least one sequence.

[0078] Specifically, generate the initial training model as shown in FIG. 4(c) according to the sequence. The constraint conditions in the figure are the design conditions, and the corresponding building design entities are also the design conditions.

[0079] In this embodiment, before obtaining the initial training model, at least one design condition of the target building needs to be obtained first. According to these design conditions, at least one design sequence can be generated, and each sequence corresponds to one or more design conditions. Finally, based on these sequences, an initial training model can be generated. This embodiment reflects the corresponding adaptation of the model based on the building entity during the model modification process, enabling the model to be better trained.

[0080] In some alternative embodiments, as shown in FIG. 5(a), step 102 includes:

[0081] Step 501, before the training, the initial training model includes the at least one design condition, the at least one design condition corresponds to a fixed value, and the user is allowed to input a fixed numerical value in the at least one design condition.

[0082] Specifically, the initial training model includes at least one design condition. In the Transformer model, one design condition corresponds to a fixed value. For example, as shown in FIG. 5(b), the association of characters corresponding to 1 is one of the key-value pairs. When the model processes Chinese characters, it will first find the vector corresponding to the Chinese character through this dictionary. This process of looking up the dictionary is called the character embedding layer.

[0083] For example, in the Transformer model, a set of all characters is first saved. For example, if there are 5000 Chinese characters in total, the size of the set is 5000. Secondly, a dictionary is saved. The dictionary stores key-value pairs with the size of the set of all characters. The keys correspond to the 5000 Chinese characters, and each key corresponds to a value, such as a three-dimensional vector. When the model processes Chinese characters, it will first find the vector corresponding to the Chinese character through this dictionary. This process of looking up the dictionary is called the character embedding layer. However, for building entities, there is no such fixed-size set. For example, for the building concept of plot length, theoretically, it can take all values from 0 to infinity, but it can only correspond to one value, that is, the classification target algorithm. Therefore, it is impossible to create a dictionary similar to Chinese characters.

[0084] Step 502, after the training, the target training model includes the at least one design condition, the at least one design condition corresponds to one or more values, and the user is allowed to input any numerical value in the at least one design condition.

[0085] Specifically, the system transforms the one-to-one structure between design conditions and values into a one-to-many structure through training. As shown in Fig. 5(c), it is then necessary to create a vector or a sequence fully-connected layer to map to a fixed-length vector as the internal representation of the model. For example, each dimension of a vector is assigned a design condition. When expressing a building entity, the building entity fills in values in the dimensions. For example, for a plot entity, there are only three values: plot length, plot width, and plot ratio. There are corresponding values in these three places, and 0 values are filled in other places. If it is a room, the length of the bedroom has a set of values, which can be any number greater than or equal to 0. That is, after training, the system changes the classification target algorithm in the transformer model to a regression target algorithm.

[0086] Step 503, after the training, the target training model includes at least one design condition including a regression loss, and the regression loss is the result calculated based on the at least one design parameter.

[0087] Specifically, the classification target algorithm usually uses a classification loss, i.e., categorical cross-entropy. However, in the trained model, a regression target algorithm is used, and it is necessary to use the mean squared error method to judge the regression loss.

[0088] When the system changes the classification target to a regression target, it is necessary to change the loss function of the model. For classification problems, categorical cross-entropy is usually used as the loss function; for regression problems, we usually use the mean squared error as the loss function. For example, as shown in the figure, the preset bedroom width is 2.5 meters and 6 meters added together, and the difference from the actual model result of 8 meters is 0.5 meters, which is the regression loss.

[0089] In this embodiment, before training, the initial training model includes at least one design condition, and the user can input a fixed positive value for optimization. After training, the target training model includes at least one design condition, allowing the user to input any value. And the target training model after training includes at least one design condition, which is associated with the regression loss. The regression loss is the result calculated based on at least one design parameter and is used to evaluate the performance of the model. This method improves the adaptability and flexibility of the model, enabling it to better meet actual needs. At the same time, it can better understand the performance of the model under different design conditions and make corresponding adjustments and optimizations to improve the prediction and decision-making ability of the model.

[0090] The further functional descriptions of the above-mentioned various modules and units are the same as those in the corresponding embodiments above, and will not be repeated here.

[0091] In this embodiment, a deep learning-based building design device is also provided. This device is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0092] This embodiment provides a deep learning-based building design device, as Figure 6 shown, including:

[0093] A first acquisition module 601, configured to acquire an initial training model and building model data to be trained, where the initial training model includes at least one design condition;

[0094] A training module 602, configured to input the building model data to be trained into the at least one design condition of the initial training model for training to obtain a target training model;

[0095] A second acquisition module 603, configured to acquire at least one design parameter of a target building, where each design parameter is generated after a user assigns values to at least one design condition of the target building;

[0096] A generation module 604, configured to input the at least one design parameter into the target training model for training, and after designing through the at least one design condition, generate a target building model corresponding to the target building.

[0097] In some optional implementation manners, the first acquisition module 601 is further specifically configured to receive a modification operation of the user on a part of the model content in the building model; acquire at least one second design parameter of the building model, where the at least one second design parameter is a parameter corresponding to the remaining unchanged part of the building model; and acquire at least one design condition of the target building.

[0098] In some optional implementation manners, the training module 602 is further specifically configured to, in response to the modification operation of the user, modify the design parameters corresponding to the building model to generate at least one first design parameter; and sort the second design parameters to generate a second design sequence.

[0099] In some alternative embodiments, the generating module 604 is further specifically configured to adaptively modify the remaining unmodified part of the building model according to the at least one first design parameter to generate a second building model of the target building; modify the building model in the order of the second design sequence to generate a second building model of the target building; arrange each of the design parameters in a preset order to generate at least one design sequence. Generate at least one design sequence according to the at least one design condition, each of the sequences corresponding to one or more design conditions; generate the initial training model according to the at least one sequence; after the training, the target training model includes at least one design condition including a regression loss, and the regression loss is a result calculated based on the at least one design parameter.

[0100] In some alternative embodiments, the first obtaining module 601 is further specifically configured to, before the training, the initial training model includes the at least one design condition, the at least one design condition corresponds to a fixed value, and the user is allowed to input a fixed numerical value in the at least one design condition; after the training, the target training model includes the at least one design condition, the at least one design condition corresponds to one or more values, and the user is allowed to input any numerical value in the at least one design condition.

[0101] An embodiment of the present invention further provides an electronic device having the above Figure 7 shown building design device based on deep learning.

[0102] Please refer to Figure 7 , Figure 7 which is a schematic structural diagram of an electronic device provided by an alternative embodiment of the present invention. As Figure 7 shown, the electronic device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including a high-speed interface and a low-speed interface. Each component communicates with each other using different buses and can be installed on a common main board or installed in other ways as needed. The processor can process instructions executed within the electronic device, including instructions stored in the memory or on the memory to display graphical information of the GUI on external input and output devices. In some alternative embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple electronic devices can be connected, and each device provides some necessary operations (for example, as a server array, a set of blade servers, or a multi-processor system). Figure 7 Here, one processor 10 is taken as an example.

[0103] The processor 10 may be a central processing unit, a network processor, or a combination thereof. Among them, the processor 10 may further include a hardware chip. The above-mentioned hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The above-mentioned programmable logic device may be a complex programmable logic device, a field-programmable gate array, a generic array logic, or any combination thereof.

[0104] Among them, the memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiments.

[0105] The memory 20 may include a program storage area and a data storage area. Among them, the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 20 may include a high-speed random access memory, and may further include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 may optionally include a memory remotely provided with respect to the processor 10, and these remote memories may be connected to the electronic device through a network. Examples of the above-mentioned network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0106] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk, or a solid-state drive; the memory 20 may further include a combination of the above types of memories.

[0107] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention may be implemented in hardware, firmware, or may be implemented as computer code recorded on a storage medium, or may be implemented as computer code originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and to be stored in a local storage medium, so that the method described herein may be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium may be a magnetic disk, an optical disc, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc.; further, the storage medium may further include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the method shown in the above embodiments is implemented.

[0108] Although embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A building design method based on deep learning, characterized in that, The method includes: Obtaining an initial training model and building model data to be trained, where the initial training model includes at least one design condition; Inputting the building model data to be trained into the at least one design condition of the initial training model for training to obtain a target training model; Obtaining at least one design parameter of a target building, where each design parameter is generated after a user assigns values to at least one design condition of the target building; Inputting the at least one design parameter into the target training model for training, and after designing through the at least one design condition, generating a target building model corresponding to the target building.

2. The method according to claim 1, characterized in that, After inputting each design parameter into a first training model for training to generate a building model corresponding to the target building, it further includes: Receiving a modification operation by the user on part of the model content in the building model; In response to the modification operation by the user, modifying the design parameter corresponding to the building model to generate at least one first design parameter; According to the at least one first design parameter, adaptively modifying the remaining unchanged part of the building model to generate a second building model of the target building.

3. The method according to claim 2, wherein The step of adaptively modifying the remaining unchanged part of the building model according to the at least one first design parameter to generate a second building model of the target building includes: Obtaining at least one second design parameter of the building model, where the at least one second design parameter is a parameter corresponding to the remaining unchanged part of the building model; Sorting the second design parameters to generate a second design sequence; Modifying the building model in the order of the second design sequence to generate a second building model of the target building.

4. The method according to claim 1, wherein After obtaining at least one design parameter of a target building, where each design parameter is generated after a user assigns values to at least one design condition of the target building, it further includes: Arranging each design parameter in a preset order to generate at least one design sequence.

5. The method according to claim 1, wherein Before obtaining the initial training model, it further includes: Obtaining at least one design condition of the target building; Generating at least one design sequence according to the at least one design condition, where each sequence corresponds to one or more design conditions; Generating the initial training model according to the at least one sequence.

6. The method according to claim 5, characterized in that, It further includes: Before the training, the initial training model includes the at least one design condition, the at least one design condition corresponds to a fixed value, and the user is allowed to input a fixed numerical value in the at least one design condition; After the training, the target training model includes the at least one design condition, the at least one design condition corresponds to one or more values, and the user is allowed to input any numerical value in the at least one design condition.

7. The method according to claim 6, wherein After the training, the step that the target training model includes at least one design condition further includes: After the training, the target training model includes at least one design condition including a regression loss, and the regression loss is a result calculated based on the at least one design parameter.

8. An architectural design device based on deep learning, characterized in that, The device includes: A first acquisition module, configured to acquire an initial training model and building model data to be trained, where the initial training model includes at least one design condition; A training module, configured to input the building model data to be trained into the at least one design condition of the initial training model for training to obtain a target training model; A second acquisition module, configured to acquire at least one design parameter of a target building, and each design parameter is generated after a user assigns values to at least one design condition of the target building; A generation module, configured to input the at least one design parameter into the target training model for training, and after designing by the at least one design condition, generate a target building model corresponding to the target building.

9. An electronic device, characterized in that, It includes a processor and a memory, and the memory is coupled to the processor; Computer-readable program instructions are stored on the memory, and when the computer-readable program instructions are executed by the processor, the building design method based on deep learning according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that, Computer instructions are stored on the computer-readable storage medium, and the computer instructions are used to cause a computer to execute the building design method based on deep learning according to any one of claims 1 to 7.

Images