Method, device and medium for generating building appearance renderings based on text description

Abstract

The embodiments of the present disclosure disclose a method, device and medium for generating a building appearance rendering based on text description. A specific implementation of the method includes: for each parameter indicator information, executing the following processing steps: selecting parameter indicator information corresponding to a target parameter indicator from each standard parameter indicator information group as the candidate parameter indicator information; determining the indicator information similarity between the parameter indicator information and each candidate parameter indicator information in the candidate parameter indicator information group; determining the candidate parameter indicator information corresponding to the indicator information similarity as the target parameter indicator information; determining the building appearance sub-image corresponding to the target parameter indicator information as the target building appearance sub-image; determining the target building appearance sub-image according to the parameter indicator information; and constructing a building appearance rendering according to each determined target building appearance sub-image. This implementation shortens the construction time of the building rendering.

Description

Technical Field

[0001] The present invention relates to the field of building appearance image generation, and in particular to a method, device and medium for generating building appearance effect images based on text description. Background Art

[0002] In the construction industry, before constructing a building, a rendering of the building is usually designed for users to refer to. At present, when constructing a rendering of a building, the usual method is: the designer first constructs a rendering, and then modifies it after communicating with the customer to form the final building rendering. However, this method usually has the following disadvantages: when the initial rendering does not meet the user's needs, it may take multiple rounds of communication to form the final building rendering, resulting in a long construction time for the building rendering. Summary of the invention

[0003] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0004] Some embodiments of the present disclosure propose a method for generating a building appearance rendering based on text description, a computer device, and a computer-readable storage medium to solve the technical problems mentioned in the above background technology section.

[0005] In a first aspect, some embodiments of the present disclosure provide a method for generating a building appearance rendering based on text description, the method comprising: obtaining a building description text input by a target user, wherein the building description text is a parameter descriptive text describing the building appearance rendering; converting each character data in the building description text according to a preset character format to obtain a converted building description text; extracting parameter indicators from the converted building description text according to a preset description parameter indicator table to obtain a parameter indicator information group; in response to determining that the target user is a historical archive user, obtaining at least one historical building appearance rendering corresponding to the target user; determining at least one standard building description text corresponding to the at least one historical building appearance rendering, wherein one historical building appearance rendering corresponds to one standard building description text, and the standard building description text includes a standard parameter indicator information group; for each parameter indicator information in the parameter indicator information group, performing the following Processing steps: select parameter indicator information corresponding to the target parameter indicator from each standard parameter indicator information group as alternative parameter indicator information to obtain an alternative parameter indicator information group, wherein the above-mentioned target parameter indicator represents the parameter indicator referred to by the above-mentioned parameter indicator information; determine the indicator information similarity between the above-mentioned parameter indicator information and each alternative parameter indicator information in the above-mentioned alternative parameter indicator information group to obtain an indicator information similarity group; in response to determining that there is an indicator information similarity that meets a first preset condition in the above-mentioned indicator information similarity group, determine the alternative parameter indicator information corresponding to the above-mentioned indicator information similarity as the target parameter indicator information; determine the building appearance sub-image corresponding to the above-mentioned target parameter indicator information as the target building appearance sub-image; in response to determining that there is no indicator information similarity that meets the first preset condition in the above-mentioned indicator information similarity group, determine the target building appearance sub-image according to the above-mentioned parameter indicator information; and construct a building appearance rendering according to each determined target building appearance sub-image.

[0006] In a second aspect, the present disclosure further provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the method described in any implementation of the first aspect is implemented.

[0007] In a third aspect, the present disclosure further provides a computer-readable storage medium, on which a computer program is stored, wherein when the computer program is executed by a processor, the method described in any implementation manner of the first aspect is implemented.

[0008] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the method for generating a building appearance rendering based on text description of some embodiments of the present disclosure, the construction time of the building rendering is shortened. First, the building description text input by the target user is obtained. The above-mentioned building description text is a parameter descriptive text describing the building appearance rendering. Thus, the building appearance rendering can be constructed according to the direct needs of the user. Secondly, according to the preset character format, each character data in the above-mentioned building description text is converted to obtain a converted building description text. Thus, it is convenient to have the same data format. Thirdly, according to the preset description parameter indicator table, the parameter indicator of the above-mentioned converted building description text is extracted to obtain a parameter indicator information group. Thus, each parameter indicator can be extracted to facilitate the determination of each component of the building rendering. Then, in response to determining that the above-mentioned target user is a historical archive user, at least one historical building appearance rendering corresponding to the above-mentioned target user is obtained. Thus, the user's preference for architectural style can be determined based on the user's historical building appearance rendering. Then, at least one standard building description text corresponding to the above-mentioned at least one historical building appearance rendering is determined. Among them, one historical building appearance rendering corresponds to one standard building description text, and the standard building description text includes a standard parameter indicator information group. Thus, the standard building description text for history is convenient to determine the building appearance sub-graph corresponding to the parameter index. Then, for each parameter index information in the above parameter index information group, the following processing steps are performed: the parameter index information corresponding to the target parameter index is selected from each standard parameter index information group as the alternative parameter index information to obtain the alternative parameter index information group, wherein the target parameter index represents the parameter index referred to by the above parameter index information; the index information similarity between the above parameter index information and each alternative parameter index information in the above alternative parameter index information group is determined to obtain the index information similarity group; in response to determining that there is an index information similarity that meets the first preset condition in the above index information similarity group, the alternative parameter index information corresponding to the above index information similarity is determined as the target parameter index information; the building appearance sub-graph corresponding to the above target parameter index information is determined as the target building appearance sub-graph; in response to determining that there is no index information similarity that meets the first preset condition in the above index information similarity group, the target building appearance sub-graph is determined according to the above parameter index information. Thus, the building appearance sub-graph corresponding to the current parameter index can be selected according to the standard building description text for history. Finally, the building appearance rendering is constructed according to each determined target building appearance sub-image. Thus, the building appearance rendering can be quickly constructed according to the building description text provided by the user and the corresponding historical building appearance rendering. Thus, the construction time of the building rendering is shortened. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The above and other features, advantages and aspects of the embodiments of the present invention will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0010] Figure 1 is a flowchart of some embodiments of a method for generating a building appearance rendering based on text description according to the present disclosure;

[0011] Figure 2 It is a schematic block diagram of the structure of a computer device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0012] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0013] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0014] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0015] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0016] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0017] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0018] Figure 1The flowchart of some embodiments of the method for generating a building appearance rendering based on text description according to the present disclosure is shown. The flowchart 100 of some embodiments of the method for generating a building appearance rendering based on text description according to the present disclosure is shown. The method for generating a building appearance rendering based on text description includes the following steps:

[0019] Step 101: Obtain the building description text input by the target user.

[0020] In some embodiments, the execution subject (e.g., a computing device) of the method for generating a building appearance rendering based on text description may obtain a building description text input by a target user. The building description text is a parameter descriptive text describing the building appearance rendering. The building description text may refer to a requirement text describing the building appearance rendering, which may include: size, style, door and window requirements, door and window style, floor height requirements, exterior wall color requirements, step style, etc. The building appearance rendering may be a rendering representing a certain type of building. For example, the building appearance rendering may be a rendering of a building to be built, or a rendering of a palace to be built.

[0021] Step 102: convert each character data in the building description text according to a preset character format to obtain a converted building description text.

[0022] In some embodiments, the execution subject may convert each character data in the building description text according to a preset character format to obtain a converted building description text. The preset character format may be a Chinese character format or an Arabic character format. For example, all parameters included in each character data in the building description text may be converted into parameters in a Chinese character format, or all parameters may be converted into parameters in an Arabic character format.

[0023] Step 103: extract parameter indicators from the converted building description text according to a preset description parameter indicator table to obtain a parameter indicator information group.

[0024] In some embodiments, the execution subject may extract parameter indicators from the converted building description text according to a preset description parameter indicator table to obtain a parameter indicator information group. The description parameter indicator table may be the various description parameter indicators of the pre-set building appearance rendering. Each description parameter indicator may refer to the description parameter indicator of each building component that constitutes the building appearance rendering. Building components may include but are not limited to: windows, doors, roofs, various rooms, steps and other components. The description parameter indicator may represent the size, style and other descriptive parameters of a certain building component. That is, the execution subject may extract the corresponding parameter indicators in the converted building description text according to the description form of the description parameter indicators in the description parameter indicator table to obtain a parameter indicator information group.

[0025] Step 104: in response to determining that the target user is a historical archive user, obtaining at least one historical building exterior rendering corresponding to the target user.

[0026] In some embodiments, the execution subject may obtain at least one historical building appearance rendering corresponding to the target user in response to determining that the target user is a historical archive user. The historical archive user may refer to a user who has constructed a building appearance rendering. That is, the execution subject may obtain at least one historical building appearance rendering corresponding to the target user from a local database.

[0027] Step 105: Determine at least one standard building description text corresponding to the at least one historical building appearance rendering.

[0028] In some embodiments, the execution subject may determine at least one standard building description text corresponding to the at least one historical building appearance rendering. One historical building appearance rendering corresponds to one standard building description text, and the standard building description text includes a standard parameter indicator information group. The standard building description text may refer to the building description text corresponding to the historical building appearance rendering, that is, the description in the standard building description text is completely consistent with the historical building appearance rendering. One standard parameter indicator information corresponds to one building component.

[0029] It should be noted that one building component corresponds to one building appearance sub-image. One building appearance rendering is composed of multiple building appearance sub-images. The standard building description text is the text formed by describing each building component in the building appearance rendering after the building appearance rendering is constructed.

[0030] Step 106: for each parameter indicator information in the above parameter indicator information group, perform the following processing steps:

[0031] Step 1061 , selecting parameter indicator information corresponding to the target parameter indicator from each standard parameter indicator information group as candidate parameter indicator information to obtain a candidate parameter indicator information group.

[0032] In some embodiments, the execution subject may select parameter indicator information corresponding to the target parameter indicator from each standard parameter indicator information group as the candidate parameter indicator information to obtain the candidate parameter indicator information group. The target parameter indicator represents the parameter indicator indicated by the parameter indicator information. That is, the parameter indicator pointed to by the candidate parameter indicator information group is the same as the parameter indicator indicated by the parameter indicator information.

[0033] Step 1062: determine the similarity between the parameter indicator information and each candidate parameter indicator information in the candidate parameter indicator information group to obtain an indicator information similarity group.

[0034] In some embodiments, the execution subject may determine the similarity between the parameter index information and each candidate parameter index information in the candidate parameter index information group to obtain an index information similarity group. That is, the cosine similarity function may be used to determine the similarity between the parameter index information and each candidate parameter index information in the candidate parameter index information group to obtain an index information similarity group.

[0035] Step 1063 , in response to determining that there is an indicator information similarity satisfying the first preset condition in the indicator information similarity group, determining the candidate parameter indicator information corresponding to the indicator information similarity as the target parameter indicator information.

[0036] In some embodiments, the execution subject may determine the candidate parameter indicator information corresponding to the indicator information similarity as the target parameter indicator information in response to determining that there is an indicator information similarity that satisfies the first preset condition in the indicator information similarity group. The first preset condition may be: the indicator information similarity is greater than or equal to the preset similarity and is the largest indicator information similarity in the indicator information similarity group; for example, the preset similarity may be 0.95.

[0037] Step 1064: determine the building appearance sub-image corresponding to the above target parameter indicator information as the target building appearance sub-image.

[0038] In some embodiments, the execution subject may determine the building appearance sub-image corresponding to the target parameter indicator information as the target building appearance sub-image, that is, the building appearance sub-image pointed to by the target parameter indicator information may be determined as the target building appearance sub-image.

[0039] Step 1065 , in response to determining that there is no indicator information similarity satisfying the first preset condition in the indicator information similarity group, determine a target building appearance sub-image according to the parameter indicator information.

[0040] In some embodiments, the execution subject may determine the target building appearance sub-image according to the parameter indicator information in response to determining that there is no indicator information similarity satisfying the first preset condition in the indicator information similarity group.

[0041] In an actual application scenario, the above execution entity can determine the target building appearance sub-image through the following steps:

[0042] The first step is to extract the parameter index feature description information and the index feature information corresponding to the above parameter index information. The parameter index feature description information can represent the description information of a certain building component, and may include: style features, color features, texture features and other descriptions. The index feature information can represent the parameter size description of a certain building component, and may include: length, width, height, color value range and other parameter values.

[0043] In the second step, the parameter indicator feature description information and the indicator feature information are input into a pre-trained building appearance sub-image association feature recognition model to obtain parameter indicator recommended feature association information. The building appearance sub-image association feature recognition model can be a pre-trained neural network model for identifying the parameter indicator feature description information and the parameter indicator recommended feature association information of the corresponding building appearance sub-image in the indicator feature information. For example, the building appearance sub-image association feature recognition model can be a plurality of fully connected modules connected in series. Each fully connected module includes: at least one fully connected layer connected in series. The parameter indicator recommended feature association information can be information representing the association relationship between each indicator feature.

[0044] The third step is to select a recommended indicator feature group from the indicator feature set corresponding to the indicator feature information according to the above parameter indicator recommendation feature association information. Among them, the indicator features in the indicator feature set may be features that the target user is interested in. For example, the indicator feature set may include: size features, style features, and color features. The indicator feature set is a feature set of the recommended indicator information set for the target user. The recommended indicator information in the recommended indicator information set may be recommended content recommended to the target user. In practice, the recommended information may be a building component diagram (building appearance sub-diagram) corresponding to a certain indicator feature. For example, a preset recommendation feature extraction method may be used to filter out a recommended indicator feature group from the indicator feature set corresponding to the above indicator feature information according to the above parameter indicator recommendation feature association information. For example, the recommendation feature extraction method may be to extract features whose corresponding connection value mean is greater than the target value as recommended indicator features.

[0045] The fourth step is to select a group of alternative building appearance sub-images from the associated building appearance sub-image database based on the above-mentioned recommended indicator feature group, and send the above-mentioned group of alternative building appearance sub-images to the user terminal of the above-mentioned target user. Each alternative building appearance sub-image corresponds to multiple indicator features. First, determine the similarity between the multiple indicator features corresponding to each of the above-mentioned alternative building appearance sub-images and the above-mentioned recommended indicator feature group. Secondly, the building appearance sub-images whose corresponding similarities are greater than the preset similarity can be determined as alternative building appearance sub-images to obtain an alternative building appearance sub-image group. The similarity can be determined by the cosine similarity formula.

[0046] In step 5, in response to receiving the selection information corresponding to the above-mentioned candidate building appearance sub-image group sent by the above-mentioned user terminal, determine the target building appearance sub-image according to the above-mentioned selection information. The selection information can be click information of the user terminal selecting a certain candidate building appearance sub-image. That is, the candidate building appearance sub-image pointed to by the selection information can be determined as the target building appearance sub-image.

[0047] Furthermore, before inputting the parameter index feature description information and the indicator feature information into a pre-trained building appearance sub-image associated feature recognition model to obtain parameter index recommended feature associated information, the method further includes:

[0048] The first step is to obtain a training sample set. The training sample includes: historical parameter indicator feature description information, historical indicator feature information and historical indicator feature set. The historical parameter indicator feature description information may refer to historical parameter indicator feature description information. The historical indicator feature information may refer to historical indicator feature information. The historical indicator feature set may be a historical indicator feature set.

[0049] In the second step, for the target training sample in the training sample set, perform the following training steps:

[0050] The first training step is to perform feature grouping on the indicator feature set corresponding to the target historical indicator feature information to generate at least one indicator feature group. The target training sample includes: target historical parameter indicator feature description information, target historical indicator feature information and target historical indicator feature set. For example, the above-mentioned execution subject can perform feature grouping on the indicator feature set corresponding to the target historical indicator feature information to generate at least one indicator feature group. That is, each indicator feature in each indicator feature group belongs to the same feature category. The feature and feature category association relationship table can be used to perform feature grouping on the indicator feature set corresponding to the target historical indicator feature information to perform feature grouping on the indicator feature set to generate at least one indicator feature group by grouping indicator features with the same feature category.

[0051] In the second training step, at least one indicator feature information group corresponding to at least one indicator feature group is determined based on the target historical indicator feature information and the target historical parameter indicator feature description information. There is a one-to-one correspondence between the indicator features in at least one indicator feature group and the indicator feature information in at least one indicator feature information group. The indicator feature information is the specific content of the feature.

[0052] As an example, the above-mentioned execution entity can query the indicator feature information corresponding to each indicator feature in at least one indicator feature group from the target historical parameter indicator feature description information and the target historical indicator feature information through information query to obtain at least one indicator feature information group.

[0053] The third training step is to input at least one indicator feature information group into the initial indicator feature cross-combination model to generate feature cross-combination indicator information. The initial building appearance sub-image associated feature recognition model includes: an initial indicator feature cross-combination model.

[0054] Among them, the feature cross-combination index information can represent the feature cross-information between each indicator feature information in at least one indicator feature information group. The initial indicator feature cross-combination model can be an indicator feature cross-combination model that has not yet been trained. The indicator feature cross-combination model can be a neural network model that cross-combines each indicator feature information. For example, the indicator feature cross-combination model can be a multi-layer fully connected layer. Among them, the initial building appearance sub-graph association feature recognition model includes: an initial indicator feature cross-combination model. The initial building appearance sub-graph association feature recognition model can be a building appearance sub-graph association feature recognition model that has not yet been trained. The building appearance sub-graph association feature can be information that represents the association relationship between each historical indicator feature of the historical indicator feature set. In practice, the historical indicator feature association relationship can be the association parameter value between each historical indicator feature. The larger the association parameter value, the higher the feature closeness between two connected historical indicator features. For example, the building appearance sub-graph association feature recognition model can be a plurality of fully connected modules connected in series. Each fully connected module includes: at least one fully connected layer connected in series. In practice, the feature cross-combination index information can be information in vector form.

[0055] The fourth training step is to determine whether the training of the initial building appearance sub-image associated feature recognition model is completed based on the feature cross-combination indicator information and the target historical indicator feature set. First, the above-mentioned execution entity can input the feature cross-combination indicator information into the remaining fully connected modules included in the initial building appearance sub-image associated feature recognition model to generate initial historical indicator feature association information. Then, feature analysis is performed on the above-mentioned initial historical indicator feature association information to generate an indicator feature set. Next, based on the above-mentioned indicator feature set and the target historical indicator feature set, indicator feature loss information is generated. Finally, in response to determining that the indicator feature loss information is less than the target loss value, it is determined that the training of the initial building appearance sub-image associated feature recognition model is completed. In response to determining that the indicator feature loss information is greater than or equal to the target loss value, it is determined that the training of the initial building appearance sub-image associated feature recognition model is not completed.

[0056] The fourth training step may include:

[0057] 1. Input the above-mentioned target historical parameter indicator feature description information and the above-mentioned target historical indicator feature information into the initial associated feature linear combination model included in the initial building appearance sub-image associated feature recognition model to obtain feature linear combination indicator information. Among them, the initial associated feature linear combination model can be an associated feature linear combination model that has not yet been trained. The associated feature linear combination model can be a neural network model that linearly combines each indicator feature information. The feature linear combination indicator information can be a linear combination of each indicator feature information between the target historical parameter indicator feature description information and the above-mentioned target historical indicator feature information. For example, the associated feature linear combination model can be a fully connected layer. In practice, the feature linear combination indicator information can be information in vector form.

[0058] 2. Generate initial indicator feature association information based on the above-mentioned feature linear combination indicator information and the above-mentioned feature cross combination indicator information. First, fuse the above-mentioned feature linear combination indicator information and the above-mentioned feature cross combination indicator information to generate fused information. Then, input the fused information into the fully connected layer to generate initial indicator feature association information.

[0059] 3. Determine whether the training of the initial building appearance sub-image associated feature recognition model is completed based on the above-mentioned initial indicator feature association information and the above-mentioned target historical indicator feature set. First, determine the indicator feature set loss information for the above-mentioned initial indicator feature association information and the above-mentioned target historical indicator feature set. Then, in response to determining that the indicator feature set loss information is less than the target loss value, determine that the training of the initial building appearance sub-image associated feature recognition model is completed. In response to determining that the indicator feature set loss information is greater than or equal to the target loss value, determine that the training of the initial building appearance sub-image associated feature recognition model is not completed.

[0060] In a fifth training step, in response to determining that the training is completed, the initial building appearance sub-image associated feature recognition model is determined as the building appearance sub-image associated feature recognition model, wherein the building appearance sub-image associated feature recognition model may be a trained neural network model.

[0061] Further, in response to determining that the training is not completed, model parameters of the initial building appearance sub-image associated feature recognition model are updated to obtain an updated building appearance sub-image associated feature recognition model.

[0062] In some embodiments, the execution subject may update the model parameters of the initial building appearance sub-image associated feature recognition model in response to determining that the training is not completed, to obtain an updated building appearance sub-image associated feature recognition model. The model parameters of the initial building appearance sub-image associated feature recognition model may be updated using algorithms such as gradient descent to obtain an updated building appearance sub-image associated feature recognition model.

[0063] Furthermore, training samples are re-extracted from the training sample set after the target training sample is removed as candidate training samples.

[0064] In some embodiments, the execution subject may re-extract training samples from the training sample set after removing the target training sample as candidate training samples. Training samples may be re-randomly extracted from the training sample set after the target training sample as candidate training samples.

[0065] Furthermore, the updated building appearance sub-image associated feature recognition model is used as the initial building appearance sub-image associated feature recognition model, and the candidate training samples are used as target training samples, and the above training steps are performed again.

[0066] In some embodiments, the execution subject may use the updated building appearance sub-image associated feature recognition model as the initial building appearance sub-image associated feature recognition model and the candidate training samples as the target training samples to perform the training steps again.

[0067] Step 107: construct a building appearance rendering according to each determined target building appearance sub-image.

[0068] In some embodiments, the execution subject may construct a building appearance rendering according to each determined target building appearance sub-image. For example, each target building appearance sub-image may be input into an image synthesizer according to requirements to obtain a building appearance rendering. For example, the requirements may be requirements recorded by the target user in the building description text, which may represent layout requirements. The image synthesizer may be a synthesizer that combines multiple images into one image.

[0069] In an actual application scenario, the above execution entity can construct a building appearance rendering through the following steps:

[0070] The first step is to input the above-mentioned target building appearance sub-images into the image synthesizer to obtain the initial building appearance renderings.

[0071] In the second step, the initial building appearance rendering is input into a pre-trained high-definition image generation model to obtain a high-definition initial building appearance rendering as the building appearance rendering. The high-definition image generation model can be a pre-trained neural network model that takes the initial building appearance rendering as input and the high-definition initial building appearance rendering as output. For example, the high-definition image generation model can be a trained Diffusion Model diffusion model.

[0072] Furthermore, the high-definition image generation model can be trained through the following steps:

[0073] The first step is to obtain a building appearance rendering training sample set, wherein the building appearance rendering training samples in the building appearance rendering training sample set include building appearance rendering sample images.

[0074] The second step is to select a target building appearance rendering training sample from the above building appearance rendering training sample set. A building appearance rendering training sample can be randomly selected from the above building appearance rendering training sample set as the target building appearance rendering training sample.

[0075] In the third step, the sample images of building appearance effects included in the training samples of the above-mentioned target building appearance effect map are input into the feature extraction network included in the initial high-definition image generation model to perform hierarchical feature extraction processing on the sample images of the above-mentioned building appearance effects to obtain the features of the sample images of the building appearance effects on each layer. The feature extraction network can be a convolutional neural network based on a deep learning method. As an example, the feature extraction network can be a backbone network that is improved by adding an attention mechanism on the basis of the backbone feature extraction network used by the target detection algorithm. For example, on the basis of the backbone network (CSPDarknet), an improved SE-CSPDarknet backbone network is obtained by adding the SE (Squeeze and Excitation) attention mechanism. The feature extraction network can include multiple convolution modules, multiple attention mechanism convolution modules, and pooling modules. It can be used to extract features of sample images of building appearance effects on each layer.

[0076] In the fourth step, the sample image features of the building appearance effects of each layer are input into the feature fusion network included in the initial high-definition image generation model to obtain the fusion features of each sample appearance effect image. The feature fusion network can be set according to actual needs, such as using a bottom-up upsampling feature fusion structure or a top-down downsampling feature fusion structure.

[0077] In the fifth step, the fusion features of the above-mentioned sample appearance effect images are input into the positioning detection network included in the above-mentioned initial high-definition image generation model to obtain the positioning detection results. Among them, the positioning detection network may include two parts: a classification structure and a positioning structure. The execution subject may input the fusion features of each sample appearance effect image into the positioning detection network in the initial high-definition image generation model, thereby obtaining the classification data and positioning data of the fusion features of each sample appearance effect image. Then, the classification data of the fusion features of each sample appearance effect image are dimensionally superimposed to obtain the total classification prediction data of the sample image of the building appearance effect. Among them, the classification prediction dimension may include the number of predicted detection frames and the predicted value of each detection frame belonging to each labeled category. In addition, the positioning data of the fusion features of each sample appearance effect image may be dimensionally superimposed to obtain the total positioning prediction data of the sample image of the building appearance effect. Among them, the positioning prediction dimension may include the number of predicted detection frames and the predicted clarity of each detection frame in the sample image.

[0078] Step 6: Based on the positioning detection results, the output result of the initial high-definition image generation model is generated. For the positioning detection results of the fusion features of each sample appearance effect image of the same sample image, the execution subject can calculate the average value as the output result of the model.

[0079] Step 7: Determine the loss value between the sample label corresponding to the target building appearance rendering training sample and the output result. In practice, the execution subject may determine the loss value between the sample label corresponding to the target building appearance rendering training sample and the output result through a preset loss function. The sample label may be a preset clarity label corresponding to the building appearance effect sample image included in the target building appearance rendering training sample.

[0080] In the eighth step, in response to determining that the loss value is less than or equal to the preset loss value, the initial high-definition image generation model is determined as the trained high-definition image generation model.

[0081] Figure 2 A schematic block diagram of the structure of a computer device provided in an embodiment of the present disclosure. The computer device may be a terminal.

[0082] like Figure 2 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0083] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any method for generating a building appearance rendering based on a text description.

[0084] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0085] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any method for generating a building appearance rendering based on text description.

[0086] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will appreciate that Figure 2 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present disclosure, and does not constitute a limitation on the computer device to which the scheme of the present disclosure is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0087] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0088] In one embodiment, the processor is used to run a computer program stored in a memory to implement the following steps: obtaining a building description text input by a target user, wherein the building description text is a parameter descriptive text describing a building appearance rendering; converting each character data in the building description text according to a preset character format to obtain a converted building description text; extracting parameter indicators from the converted building description text according to a preset description parameter indicator table to obtain a parameter indicator information group; in response to determining that the target user is a historical archive user, obtaining at least one historical building appearance rendering corresponding to the target user; determining at least one standard building description text corresponding to the at least one historical building appearance rendering, wherein one historical building appearance rendering corresponds to one standard building description text, and the standard building description text includes a standard parameter indicator information group; for each parameter indicator information in the parameter indicator information group, executing the following steps Processing steps: selecting parameter indicator information corresponding to the target parameter indicator from each standard parameter indicator information group as the alternative parameter indicator information to obtain the alternative parameter indicator information group, wherein the above-mentioned target parameter indicator represents the parameter indicator referred to by the above-mentioned parameter indicator information; determining the indicator information similarity between the above-mentioned parameter indicator information and each alternative parameter indicator information in the above-mentioned alternative parameter indicator information group to obtain an indicator information similarity group; in response to determining that there is an indicator information similarity that meets a first preset condition in the above-mentioned indicator information similarity group, determining the alternative parameter indicator information corresponding to the above-mentioned indicator information similarity as the target parameter indicator information; determining the building appearance sub-image corresponding to the above-mentioned target parameter indicator information as the target building appearance sub-image; in response to determining that there is no indicator information similarity that meets the first preset condition in the above-mentioned indicator information similarity group, determining the target building appearance sub-image according to the above-mentioned parameter indicator information; and constructing a building appearance rendering according to each determined target building appearance sub-image.

[0089] The embodiments of the present disclosure also provide a computer-readable storage medium, on which a computer program is stored. The computer program includes program instructions. The method implemented when the program instructions are executed can refer to the various embodiments of the method for generating a building appearance rendering based on text description in the present disclosure.

[0090] The computer-readable storage medium may be an internal storage unit of the computer device in the aforementioned embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SmartMediaCard, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., provided on the computer device.

[0091] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0092] The serial numbers of the embodiments of the present disclosure are only for description and do not represent the advantages and disadvantages of the embodiments. The above description is only a specific implementation mode of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present disclosure, and these modifications or replacements should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be based on the protection scope of the claims.

Claims

1. A method for generating a building appearance rendering based on text description, characterized in that: The method comprises: Acquire a building description text input by a target user, wherein the building description text is a parameter descriptive text describing a building appearance rendering; Converting each character data in the building description text according to a preset character format to obtain a converted building description text; According to a preset description parameter index table, extracting parameter indexes from the converted building description text to obtain a parameter index information group; In response to determining that the target user is a historical archive user, obtaining at least one historical building appearance rendering corresponding to the target user; Determine at least one standard building description text corresponding to the at least one historical building appearance rendering, wherein one historical building appearance rendering corresponds to one standard building description text, and the standard building description text includes a standard parameter indicator information group; For each parameter indicator information in the parameter indicator information group, the following processing steps are performed: Selecting parameter indicator information corresponding to the target parameter indicator from each standard parameter indicator information group as candidate parameter indicator information to obtain a candidate parameter indicator information group, wherein the target parameter indicator represents the parameter indicator indicated by the parameter indicator information; Determine the similarity of the parameter index information with each candidate parameter index information in the candidate parameter index information group to obtain an index information similarity group; In response to determining that there is an indicator information similarity satisfying a first preset condition in the indicator information similarity group, determining the candidate parameter indicator information corresponding to the indicator information similarity as the target parameter indicator information; Determine the building appearance sub-image corresponding to the target parameter index information as the target building appearance sub-image; In response to determining that there is no indicator information similarity satisfying a first preset condition in the indicator information similarity group, determining a target building appearance sub-graph according to the parameter indicator information; Constructing a building appearance rendering according to each determined target building appearance sub-image; The step of determining the target building appearance sub-image according to the parameter indicator information includes: Extracting parameter indicator feature description information and indicator feature information corresponding to the parameter indicator information; Acquire a training sample set, wherein the training sample includes: historical parameter indicator feature description information, historical indicator feature information and historical indicator feature set; For the target training sample in the training sample set, perform the following training steps: Performing feature grouping on the target historical indicator feature information corresponding to the indicator feature set to generate at least one indicator feature group, wherein the target training sample includes: target historical parameter indicator feature description information, target historical indicator feature information and target historical indicator feature set; Determine at least one indicator feature information group corresponding to at least one indicator feature group according to the target historical indicator feature information and the target historical parameter indicator feature description information; Inputting at least one indicator feature information group into an initial indicator feature cross-combination model to generate feature cross-combination indicator information, wherein the initial building appearance sub-image associated feature recognition model comprises: an initial indicator feature cross-combination model; Determine whether the initial building appearance sub-image associated feature recognition model has been trained based on the feature cross-combination indicator information and the target historical indicator feature set; In response to determining that the training is completed, determining the initial building appearance sub-image associated feature recognition model as the building appearance sub-image associated feature recognition model; Inputting the parameter index feature description information and the index feature information into a pre-trained building appearance sub-image associated feature recognition model to obtain parameter index recommended feature associated information; According to the parameter indicator recommended feature association information, a recommended indicator feature group is selected from the indicator feature set corresponding to the indicator feature information; According to the recommendation indicator feature group, selecting a candidate building appearance sub-image group from a related building appearance sub-image database, and sending the candidate building appearance sub-image group to a user terminal of the target user; In response to receiving the selection information corresponding to the candidate building appearance sub-image group sent by the user terminal, determining a target building appearance sub-image according to the selection information; Wherein, determining whether the initial building appearance sub-image associated feature recognition model has been trained includes: inputting the above-mentioned target historical parameter index feature description information and the above-mentioned target historical index feature information into the initial associated feature linear combination model included in the initial building appearance sub-image associated feature recognition model to obtain feature linear combination index information; Generate initial indicator feature association information based on the above-mentioned feature linear combination indicator information and the above-mentioned feature cross-combination indicator information; Determine whether the initial building appearance sub-image associated feature recognition model has been trained based on the initial indicator feature associated information and the target historical indicator feature set; The step of constructing a building appearance rendering according to each determined target building appearance sub-image comprises: Inputting each target building appearance sub-image into an image synthesizer to obtain an initial building appearance rendering; Inputting the initial building appearance rendering into a pre-trained high-definition image generation model to obtain a high-definition initial building appearance rendering as the building appearance rendering; Among them, the high-definition image generation model is trained through the following steps: Acquire a building appearance rendering training sample set, wherein the building appearance rendering training samples in the building appearance rendering training sample set include building appearance rendering sample images; Selecting a target building appearance rendering training sample from the above building appearance rendering training sample set; Inputting the sample images of building appearance effects included in the training samples of the target building appearance effect images into the feature extraction network included in the initial high-definition image generation model to perform hierarchical feature extraction processing on the sample images of building appearance effects to obtain features of the sample images of building appearance effects at each layer; Input the sample image features of the building appearance effects of each floor into the feature fusion network included in the initial high-definition image generation model to obtain the fusion features of the sample appearance effect images; Input the fusion features of the above-mentioned sample appearance effect images into the positioning detection network included in the above-mentioned initial high-definition image generation model to obtain the positioning detection result; Based on the positioning detection results, the output results of the initial high-definition image generation model are generated; the loss value between the sample label corresponding to the training sample of the target building appearance rendering and the output results is determined; In response to determining that the loss value is less than or equal to the preset loss value, the initial high-definition image generation model is determined as a trained high-definition image generation model.

2. A computer device, wherein: The computer device comprises a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein the computer program implements the steps of the method according to claim 1 when executed by the processor.

3. A computer-readable storage medium, wherein: The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the method according to claim 1 are implemented.

Images