JP7375405B2 - Learning support system, learning support method, and learning support program - Google Patents

Description

The present invention relates to a learning support system, a learning support method, and a learning support program for creating teacher data used in machine learning for determining the construction status and progress status of construction work, etc.

At construction sites such as architecture and civil engineering, photographs may be taken before, during and after the construction work. These photos are used to manage the progress of construction and confirm the construction status. Therefore, techniques for efficiently determining the progress status of construction work, etc. are being considered (see, for example, Patent Document 1). In the technique disclosed in this document, a control unit of a support server specifies architectural elements included in a photographed image recorded in a site image information recording unit. Next, the process associated with the building element is specified using the building element information storage unit, and the specified process is output to the administrator terminal.

Furthermore, in machine learning, supervised learning using teacher data may be used. Techniques for reducing the burden of creating such training data are also being considered (see, for example, Patent Document 2). In the technology disclosed in this document, the calculation unit of the video display system has a prediction model that is machine-learned using a plurality of results of simulations of a plurality of states of a manufacturing device as training data. Then, at least a part of the first state acquired by the information acquisition unit is input into the prediction model, a second state inside the manufacturing apparatus is calculated, and video information is generated from the second state.

Japanese Patent Application Publication No. 2017-107443 JP2018-169818A

However, the materials used for construction and the size of the building differ depending on the construction site. Since the situation changes significantly depending on the progress of the construction work, it becomes a burden to prepare the teacher data. It is also possible to perform machine learning using images and simulation images for each material. However, depending on the site conditions, it is difficult to accurately judge the situation unless it is possible to distinguish between materials used for construction and materials temporarily stored.

A learning support system to solve the above problems includes a design information storage unit that records design information of a structure composed of a three-dimensional element model including architectural elements associated with each process, and a design information storage unit that is used for machine learning. It includes a teacher information storage section that records teacher data, and a control section that generates a predictive model that predicts the progress of each process. The control unit uses the design information to sequentially arrange, in a model space, three-dimensional models in which element models of architectural elements used in each process are combined, and from a viewpoint in the model space, A two-dimensional image of the model is generated and recorded in the teacher information storage section, and a prediction model that predicts the progress of each process from the photographed image is created using the teacher data recorded in the teacher information storage section. generate.

According to the present invention, it is possible to efficiently create training data used in machine learning for determining situations in construction work, etc.

FIG. 2 is an explanatory diagram of the system of this embodiment. FIG. 2 is an explanatory diagram of the hardware configuration of this embodiment. FIG. 2 is an explanatory diagram of data recorded in the storage unit of the present embodiment, in which (a) is the BIM information storage unit, (b) is the teacher information storage unit, (c) is the learning result storage unit, and (d) is the on-site An explanatory diagram of an image information storage unit and (e) a progress information storage unit. FIG. 3 is an explanatory diagram of the processing procedure of teacher data generation processing according to the present embodiment. FIG. 2 is an explanatory diagram of the processing procedure of the present embodiment, in which (a) is a learning process and (b) is an explanatory diagram of a prediction process. FIG. 2 is an explanatory diagram of the teacher data used in this embodiment, in which (a) is an image of completed wall base construction, (b) is an image of plasterboard being installed, and (c) is an image of output of material information for each part. Explanatory diagram. FIG. 7 is an explanatory diagram of the processing procedure of surface model generation processing according to another embodiment. FIG. 6 is an explanatory diagram of teacher data used in another embodiment, in which (a) is an image of materials temporarily placed at a first position, (b) is an image of materials temporarily placed at a second position, and (c) is an image of materials temporarily placed. (d) is an image where the material is temporarily placed at the first angle, (e) is an image where the material is temporarily placed at the second angle, (f) is an image where the material is temporarily placed at the third angle. An explanatory diagram of a temporarily placed image.

One embodiment will be described below with reference to FIGS. 1 to 6. This embodiment will be described as a learning support system for determining the progress status of a process at a construction site. In this embodiment, training data is created using CG images generated using BIM (Building Information Modeling). In this BIM, it is assumed that attribute data such as cost, finish, and management information can be added to a three-dimensional digital model of a building created on a computer.
In this embodiment, as shown in FIG. 1, a user terminal 10 and a support server 20 are used.

(Explanation of hardware configuration)
The hardware configuration of the information processing apparatus H10 that constitutes the user terminal 10 and the support server 20 will be explained using FIG. 2. The information processing device H10 includes a communication device H11, an input device H12, a display device H13, a storage unit H14, and a processor H15. Note that this hardware configuration is just an example, and it can also be implemented using other hardware.

The communication device H11 is an interface that establishes a communication path with other devices and executes data transmission and reception, and is, for example, a network interface card, a wireless interface, or the like.

The input device H12 is a device that accepts input of various information, and is, for example, a mouse, a keyboard, a camera (photographing device), or the like. The display device H13 is a display or the like that displays various information.
The storage unit H14 is a storage device that stores data and various programs for executing various functions of the user terminal 10 and the support server 20. Examples of the storage unit H14 include ROM, RAM, hard disk, and the like.

The processor H15 controls each process in the user terminal 10 and the support server 20 using programs and data stored in the storage unit H14. Examples of the processor H15 include a CPU, an MPU, and the like. This processor H15 expands a program stored in a ROM or the like into a RAM and executes various processes for each processing.

The processor H15 is not limited to performing software processing for all processes that it executes. For example, the processor H15 may include a dedicated hardware circuit (for example, an application-specific integrated circuit: ASIC) that performs hardware processing for at least part of the processing that it executes. That is, the processor H15 is [1] one or more processors that operate according to a computer program (software), [2] one or more dedicated hardware circuits that execute at least some of various processes, or [2] one or more dedicated hardware circuits that execute at least some of various processes. 3] Can be configured as a circuit including a combination thereof. A processor includes a CPU and memory, such as RAM and ROM, where the memory stores program codes or instructions configured to cause the CPU to perform processing. Memory or computer-readable media includes any available media that can be accessed by a general purpose or special purpose computer.

(System configuration)
Next, each function of the learning support system will be explained using FIG. 1.
The user terminal 10 is a computer terminal equipped with an imaging device used to photograph a construction site. This user terminal 10 is carried by a person in charge of the site and is used to photograph the site as necessary in order to check the progress of the construction work. The user terminal 10 then uploads the captured image to the support server 20.

The support server 20 is a computer system that supports management of the progress of construction using captured images of the construction site. The support server 20 includes a control section 21, a BIM information storage section 22 (design information storage section), a teacher information storage section 23, a learning result storage section 24, a site image information storage section 25, and a progress information storage section 26.

The control unit 21 performs processing (including a modeling stage, a teacher data generation stage, a learning stage, a prediction stage, a determination stage, etc.) to be described later. By executing programs for each process for this purpose, the control unit 21 functions as a modeling unit 210, a teacher data generation unit 211, a learning unit 212, a prediction unit 213, a determination unit 214, and the like.

The modeling unit 210 executes processing to generate a three-dimensional model of a structure using BIM data.
The teacher data generation unit 211 uses the modeling unit 210 to execute a process of generating teacher data used for machine learning.

The learning unit 212 executes a process of generating a prediction model for predicting architectural elements included in a photographed image by machine learning using teacher data.
The prediction unit 213 executes a process of predicting architectural elements included in the photographed image using the prediction model.
The determination unit 214 executes a process of determining the progress status (step) of construction using the predicted architectural elements.

As shown in FIG. 3A, BIM data 220 created in BIM is recorded in the BIM information storage unit 22. This BIM data 220 is recorded when a building is designed using three-dimensional CAD. The BIM data 220 includes a project, an element model, and a layout. Furthermore, in this embodiment, information regarding attributes, processes, and scheduled dates is included in association with the element model.

The project information includes information regarding the name of the building, longitude/latitude, architectural direction, etc.
The element model information is information regarding a three-dimensional model (BIM object) of each architectural element used in a building.
The placement information includes information regarding the coordinates at which each element model is placed.
The attribute information includes information regarding the attributes of each element model. This attribute includes element type, quantity, information regarding the architectural element, etc. Element type information includes information identifying walls, slabs, columns, beams, roofs, stairs, zones, etc. The quantity information includes information regarding length, area, volume, etc. The architectural element information includes information identifying reinforcing steel, concrete, urethane, light iron, plasterboard, floor panels, cloth, fittings, flooring, and the like. The architectural element information can display the visual color, brightness homogeneity, and texture such as unevenness of the surface of the architectural element material.

The process information includes information for specifying the process using the architectural element. This process includes, for example, "before concrete placement,""beforeinsulation,""afterinsulation,""beforeboarding,""afterboarding," and "finishing completion." The architectural elements that can be confirmed on site differ depending on the process. For example, reinforcing bars can be seen "before concrete is poured." Concrete can be seen ``before insulation'', and urethane can be seen ``after insulation''.
The scheduled date information includes information regarding the scheduled date of this process.

As shown in FIG. 3(b), teacher data 230 used for machine learning is recorded in the teacher information storage unit 23. This teacher data 230 is recorded when a teacher data creation process, which will be described later, is executed. The teacher data 230 includes data regarding outdoor and indoor perspectives (teacher images generated by two-dimensional CG) of each location and each process generated using BIM, and architectural elements included therein. be done. Here, each architectural element included in the perspective is associated with attribute information (architectural element information) included in the BIM data.

As shown in FIG. 3C, a prediction model 240 for predicting architectural elements included in a captured image is recorded in the learning result storage unit 24. This predictive model 240 is recorded when machine learning is performed using teacher data. The prediction model 240 is configured by a network (middle layer) that receives a photographed image as an input and outputs the proportion of architectural elements included in the photographed image.

As shown in FIG. 3(d), the site image information storage unit 25 records image management data 250 regarding captured images of a construction site. This image management data 250 is recorded when a captured image is acquired from the user terminal 10. The image management data 250 includes information regarding the image code, the date and time of shooting, the location, and the shot image.

The image code information includes information regarding an identifier for identifying each photographed image of a construction site.
The photographing date and time information includes information regarding the date and time when this photographed image was photographed.
The location information includes information for specifying the location of the construction site where this captured image was captured.
The photographed image information includes a photographed image of a construction site.

As shown in FIG. 3(e), progress management data 260 for managing the progress of construction work is recorded in the progress information storage unit 26. This progress management data 260 is recorded when the process determination process is executed. The progress management data 260 includes information regarding the location, image code, process, and situation.
The location information includes an identifier for identifying each location on the construction site.
The image code information includes an identifier for specifying a photographed image of this location.
The process information includes an identifier for specifying the process specified based on the photographed image. This process information further includes information regarding the ratio of architectural elements included in the photographed image.
The status information includes information indicating the status of this process with respect to the construction schedule.

Next, a processing procedure for determining the progress status of construction work in the support server 20 configured as described above will be explained using FIGS. 4 to 6.
(Teacher data creation process)
First, the teacher data creation process will be explained using FIG.
Here, the control unit 21 of the support server 20 executes a BIM data acquisition process (step S1-1). Specifically, the teacher data generation unit 211 of the control unit 21 acquires BIM data 220 of the building whose progress is to be managed from the BIM information storage unit 22.

Next, the control unit 21 of the support server 20 sequentially identifies each step of the construction work as a processing target, and repeats the following processing.
The control unit 21 of the support server 20 executes surface model generation processing (step S1-2). Specifically, the teacher data generation unit 211 of the control unit 21 specifies architectural elements used up to the process to be processed in the BIM data. Then, the teacher data generation unit 211 instructs the modeling unit 210 to create a surface model of the building using the three-dimensional models of all the identified building elements. In this case, the modeling unit 210 generates a three-dimensional model using architectural elements associated with the processes up to the process to be processed, and places it in the model space (three-dimensional space). For example, the architectural elements visible in the state of "wall base construction completed" are "concrete/LGS", and the architectural elements visible in the state of "gypsum board completed" are "concrete/gypsum board", etc.

Here, the control unit 21 of the support server 20 executes rendering condition specifying processing (step S1-3). Specifically, the teacher data generation unit 211 of the control unit 21 specifies the scheduled date (month and day) of the BIM data 220 recorded in the BIM information storage unit 22. Next, the teacher data generation unit 211 identifies a plurality of time periods in which the person in charge is likely to perform photography. The teacher data generation unit 211 specifies the sun position on the scheduled date and time, and specifies the sunlight situation for the building. In this case, an opening is identified in the surface model of the building, and sunlight from this opening is used as a light source. Furthermore, if the installation of a lighting fixture is registered in the space, this lighting fixture is included as a light source. Then, the teacher data generation unit 211 specifies a plurality of rendering conditions using the specified light source.

Next, the control unit 21 of the support server 20 sequentially uses the plurality of specified rendering conditions to execute rendering processing (step S1-4). Specifically, the modeling unit 210 of the control unit 21 executes rendering of the three-dimensional model using the rendering conditions.

Next, the control unit 21 of the support server 20 executes a viewpoint position specifying process (step S1-5). Specifically, the teacher data generation unit 211 of the control unit 21 specifies a shooting position (viewpoint) in the model space. This viewpoint is set in consideration of the position in the model space where the person in charge takes pictures. For example, a plurality of viewpoint positions are specified in the model space at positions where the person in charge can stand upright and are separated from the material by a predetermined distance, based on the height when standing upright. Note that the viewpoint position may be any position where the building materials can be seen, and is not limited to a plurality of viewpoint positions at a predetermined distance apart and at the height when standing upright.

Then, the control unit 21 of the support server 20 repeats the following process for each specified specified position.
The control unit 21 of the support server 20 executes a CG image generation process (step S1-6). Specifically, the teacher data generation section 211 of the control section 21 instructs the modeling section 210 about the viewpoint position. In this case, the modeling unit 210 generates a perspective image (CG image) at the viewpoint position. This perspective image includes each architectural element (attribute information) included in the image.

An image 500 shown in FIG. 6(a) is a perspective view generated in the process "wall base construction completion". In this image 500, a base material 501, a floor material 502, and a wall material 503 are arranged as architectural elements. Further, an image 510 shown in FIG. 6(b) is a perspective generated in the process "during plasterboard construction". In this image 510, a gypsum board 511 is arranged in addition to a base material 501, a floor material 502, and a wall material 503. The image 520 shown in FIG. 6(c) includes the architectural elements included in the image 510. Specifically, architectural element information 521-524 is set corresponding to base material 501-gypsum board 511. Then, using the area of each architectural element, the proportion of each architectural element can be calculated.

Next, the control unit 21 of the support server 20 executes a recording process of teacher data (step S1-7). Specifically, the teacher data generation section 211 of the control section 21 acquires a perspective image from the modeling section 210. Next, the teacher data generation unit 211 calculates the area ratio of each architectural element within the perspective image. Then, the teacher data generation unit 211 records teacher data in which the area ratio (progress status of the process) of each architectural element is set in association with the perspective image in the teacher information storage unit 23.
Then, the control unit 21 of the support server 20 repeats the above-described process for each viewpoint position, rendering condition, and process.

(Learning process)
Next, the learning process will be explained using FIG. 5(a).
First, the control unit 21 of the support server 20 executes teacher data calling processing (step S2-1). Specifically, the learning unit 212 of the control unit 21 acquires teacher data 230 from the teacher information storage unit 23. In this teacher data 230, a ratio is recorded for each architectural element included in the perspective image.

Next, the control unit 21 of the support server 20 executes machine learning processing (step S2-2). Here, as machine learning, deep learning is used that connects multiple nodes included in the input layer, intermediate layer (hidden layer), and output layer. Specifically, the learning unit 212 of the control unit 21 uses a perspective image for the input layer, and uses the area ratio of the architectural element for each node of the output layer. Then, a prediction model (architectural element determination model) consisting of intermediate layer nodes and connections connecting input layer nodes to output layer nodes is generated. The learning unit 212 then records the generated prediction model in the learning result storage unit 24.

(Process judgment process)
Next, the process determination process will be explained using FIG. 5(b).
First, the control unit 21 of the support server 20 executes a captured image acquisition process (step S3-1). Specifically, a construction site of construction work is photographed using the user terminal 10. In this case, the user terminal 10 generates a photographed image of the scene. The user terminal 10 then obtains the shooting date and time from the system timer of the user terminal 10. Further, the user terminal 10 acquires an identifier that specifies the location where the image was taken, by manual input using a user interface or by position specifying means such as GPS (Global Positioning System). Then, the user terminal 10 stores the generated photographed image in association with an identifier specifying the date and time of photographing and the photographing location. Further, the user terminal 10 uploads the stored captured image to the support server 20. This photographed image is uploaded with information regarding the date and time of photographing and location associated with it. The prediction unit 213 of the control unit 21 assigns an image code to the uploaded photographed image and records it in the on-site image information storage unit 25.

Next, the control unit 21 of the support server 20 sequentially identifies the captured images recorded in the on-site image information storage unit 25, and repeats the following process for each image to be processed.
First, the control unit 21 of the support server 20 executes architectural element prediction processing (step S3-2). Specifically, the prediction unit 213 of the control unit 21 inputs the image to be processed into the input layer of the prediction model, and predicts the proportion of each architectural element in the output layer.

Next, the control unit 21 of the support server 20 executes process prediction processing (step S3-3). Specifically, the determination unit 214 of the control unit 21 uses the BIM information storage unit 22 to identify an element model in which each architectural element predicted by the prediction model is recorded as an attribute. In this case, the element models of each architectural element are identified in the order of their predicted ratios. Further, the determination unit 214 searches for a process associated with the identified element model. Then, the process extracted by the search is specified. For example, if you search for processes associated with the architectural elements "concrete" and "urethane" and extract "after applying insulation material" as a process associated with these architectural elements, then this "after applying insulation material" Identify. Note that the "light railway" and "floor panels" that are not included in the photographed images are included in the process "Before board pasting" and the process "After board pasting," respectively, but are not included in "After applying insulation material." . On the other hand, if there is an excess or deficiency of architectural elements included in each process in the BIM information storage unit 22 in the photographed image, it is determined that the process cannot be specified.

Here, if it is determined that the process cannot be identified, the determining unit 214 identifies a process that includes at least one of the architectural elements included in the photographed image as a process candidate. Then, the determination unit 214 outputs the specified building candidate to the administrator's user terminal 10. In this case, process candidates are output in the order of the number of common architectural elements between the architectural elements recorded in the BIM information storage unit 22 and the architectural elements included in the photographed image. Then, the determination unit 214 acquires the process specified by the administrator via the user terminal 10 from among the predicted process candidates.

Next, the control unit 21 of the support server 20 executes a process recording process (step S3-4). Specifically, the determination unit 214 of the control unit 21 records in the progress information storage unit 26 progress management data 260 that records the proportion of architectural elements included in the photographed image together with the identified process. Further, the determination unit 214 acquires BIM data 220 in which the shooting location and process are recorded from the BIM information storage unit 22. Next, the determination unit 214 compares the scheduled date recorded in the BIM data 220 and the photographing date recorded in the site image information storage unit 25 to calculate the number of days difference. The determination unit 214 then records the situation in the progress management data 260. Here, if the shooting date is later than the scheduled completion date, a "delay in the number of days difference" is recorded in the situation data area. On the other hand, if the shooting date is earlier than the scheduled completion date or if the shooting date and the scheduled completion date match, "as scheduled" is recorded in the status data area.
The above process is repeated until completed for all captured images.

According to this embodiment, the following effects can be obtained.
(1) In the present embodiment, in the teacher data creation process, the control unit 21 of the support server 20 performs BIM data acquisition processing (step S1-1), surface model generation processing (step S1-2), and CG image generation processing. A generation process (step S1-6) and a teacher data recording process (step S1-7) are executed. Normally, the burden of preparing training data for machine learning is large. Further, even if images of individual architectural elements can be prepared as training data, it is difficult to prepare images of the construction site itself as training data. In this embodiment, by using BIM data, training data used for machine learning can be efficiently generated.

(2) In this embodiment, the control unit 21 of the support server 20 executes rendering condition specifying processing (step S1-3) and rendering processing (step S1-4). As a result, a realistic perspective image can be generated depending on the time when the process is performed, and can be used as training data.

(3) In the present embodiment, the control unit 21 of the support server 20 executes a viewpoint position specifying process (step S1-5). This makes it possible to generate training data for use in machine learning, taking into consideration the shooting situation by the person in charge.

(4) In the present embodiment, the control unit 21 of the support server 20 executes architectural element prediction processing (step S3-2) and process prediction processing (step S3-3). Thereby, the progress of the process can be determined according to the proportion of architectural elements included in the photographed image.

This embodiment can be modified and implemented as follows. This embodiment and the following modified examples can be implemented in combination with each other within a technically consistent range.
- In the above embodiment, a perspective image is generated using the BIM data 220 recorded in the BIM information storage unit 22 and used as teacher data. Elements included in this perspective image are not limited to architectural elements. For example, temporary materials placed at a construction site may be included. In this case, the BIM information storage unit 22 stores a three-dimensional model of the shape of each material used in construction when it is temporarily placed. For example, when temporarily storing plasterboard, a three-dimensional model of a shape in which a plurality of plasterboards are stacked flat is recorded.

Therefore, the generation process (step S1-2) of a surface model including temporarily placed materials will be explained using FIG. Here, a three-dimensional model of temporary storage materials is placed in the model space.
First, the control unit 21 of the support server 20 executes architectural element placement processing (step S4-1). Specifically, the teacher data generation unit 211 of the control unit 21 uses the BIM data 220 recorded in the BIM information storage unit 22 to identify architectural elements set up to the process to be processed. Then, the teacher data generation unit 211 instructs the modeling unit 210 to create a surface model of the building using the three-dimensional model of the identified architectural element.

Next, the control unit 21 of the support server 20 executes temporary storage material identification processing (step S4-2). Specifically, the teacher data generation unit 211 of the control unit 21 uses the BIM information storage unit 22 to identify materials that are likely to be temporarily stored at the time when the process to be processed is performed. For example, the process before and after the process to be processed is specified, and the architectural elements (materials) used in this process are specified as temporary materials.

Next, the control unit 21 of the support server 20 executes temporary storage material placement processing (step S4-3). Specifically, the teacher data generation unit 211 of the control unit 21 acquires a three-dimensional model of temporarily placed materials from the BIM information storage unit 22. Then, the teacher data generation unit 211 arranges the three-dimensional model of the temporary storage material at a random position on the floor in the model space of this step.

Here, as shown in FIG. 8(a), a three-dimensional model 601 of temporarily placed materials is placed at a first position with respect to the architectural element 600. Furthermore, as shown in FIGS. 8(b) and 8(c), three-dimensional models 601 of temporary materials are placed at second and third positions in the model space. Further, as shown in FIGS. 8(d), (e), and (f), the three-dimensional model 601 of the temporary storage material is rotated at any first to third angles in the model space.
In this case, a label (attribute) indicating that the material is a temporary storage material is added to the perspective image of the teacher data.

Then, in the learning process, machine learning is performed using the perspective image in which the temporarily placed materials are placed. In the process determination process, temporarily placed materials are excluded from the photographed image and the proportion of architectural elements is calculated.
As a result, even when temporarily placed materials are placed at a construction site, the progress of the process can be determined based on architectural elements, excluding the temporarily placed materials.
Furthermore, the progress may be determined using the materials of the excluded temporary materials. In this case, if temporary storage materials to be used in the next subsequent process are detected, it is determined that preparations for the subsequent process are being made.

- In the above embodiment, the user terminal 10 and the support server 20 are used. Then, the support server 20 executes process determination processing. The hardware that executes the process determination process is not limited to the support server 20. For example, the user terminal 10 may be provided with the learning result storage section 24 to the progress information storage section 26, and the process determination process may be executed in the user terminal 10.

- In the above embodiment, the control unit 21 of the support server 20 executes the learning process. In this learning process, a verification process may be executed. In this case, machine learning is performed using some of the plurality of perspective images created in the teacher data creation process. In the verification process, the control unit 21 executes a process prediction process (step S3-3) using the parsed image that is not used for machine learning. In this verification process, the control unit 21 determines the rate at which the process could not be correctly determined (error rate).

The control unit 21 records the generated prediction model in the learning result storage unit 24 when the error rate is less than the reference value. On the other hand, if the error rate is equal to or higher than the reference value, the control unit 21 re-executes the teacher data generation process, adds teacher data, and re-executes the learning process using the added teacher data.

- In the above embodiment, a perspective image is used as the teacher data. The training data is not limited to perspective images. For example, a 360-degree panoramic image in all directions, up, down, left, and right may be used. In this case, a panoramic image from the viewpoint is generated in the model space. When photographing the scene, a spherical camera capable of capturing panoramic images is used.
Furthermore, the teacher data is not limited to CG images using BIM data. For example, an image (Mixed Reality image, etc.) that is a combination of a real image and a CG image may be used. In this case, a real image can be used as the background, or an image in which a real architectural element image is superimposed on the BIM model can be used.

- In the above embodiment, the control unit 21 of the support server 20 executes machine learning processing (step S2-2). Here, a perspective image is used for the input layer, and the area ratio of architectural elements is used for each node of the output layer as information indicating the progress of the process. Information that allows the process to be predicted may be set in the output layer. For example, a perspective image may be used for the input layer, and architectural elements may be used for each node of the output layer. In this case, the progress of the process is estimated based on the combination of architectural elements predicted using the photographed images.

- In the above embodiment, the control unit 21 of the support server 20 executes rendering condition specifying processing (step S1-3). Here, the rendering conditions are specified using the scheduled date and time zone. The rendering conditions are not limited to using the scheduled date and time zone. For example, an image may be used that has undergone image processing to reduce the influence of the light source and characterize visible building materials. In this case, the same image processing is performed on the photographed image during the process determination process.

- In the above embodiment, the control unit 21 of the support server 20 executes rendering condition specifying processing (step S1-3). Here, the surrounding environment of the construction site may be considered. In this case, the BIM information storage unit 22 stores three-dimensional models of surrounding buildings. When generating a perspective image, the arrangement of light sources is determined in consideration of surrounding buildings. Further, surrounding buildings visible through the opening of the structure may be included in the perspective image as information other than architectural elements. Thereby, machine learning can be performed using a CG image that is close to reality.

- In the above embodiment, the control unit 21 of the support server 20 executes the teacher data recording process (step S1-7) and the machine learning process (step S2-2). In the process determination process, the control unit 21 of the support server 20 executes architectural element prediction processing (step S3-2). Here, the teacher data generation unit 211 performs machine learning using teacher data in which the area ratio of each architectural element is set in association with the perspective image. The training data is not limited to the case where area ratios are used. For example, the process may be predicted directly from the image. In this case, machine learning is performed using training data in which process information is set in association with the perspective image. Then, a predictive model is generated using a perspective image in the input layer and process information in the output layer. In this case, the control unit 21 of the support server 20 executes process prediction processing using this prediction model (step S3-3).

- In the above embodiment, the control unit 21 of the support server 20 executes a process recording process (step S3-4). Here, the scheduled date recorded in the BIM data 220 and the photographing date recorded in the site image information storage section 25 are compared to calculate the number of days difference, and the situation is recorded in the progress management data 260. The recorded content of the process is not limited to this. For example, if the scheduled date is not recorded in the BIM data 220, process information is recorded in association with the photographing date.

DESCRIPTION OF SYMBOLS 10... User terminal, 20... Support server, 21... Control unit, 210... Modeling unit, 211... Teacher data generation unit, 212... Learning unit, 213... Prediction unit, 214... Process determination unit, 22... BIM information storage unit, 23...Teacher information storage unit, 24...Learning result storage unit, 25...Site image information storage unit, 26...Progress information storage unit.

Claims (9)

a design information storage unit that records design information of a structure configured by a three-dimensional element model including architectural elements associated with each process of construction work ;
a teacher information storage unit that records teacher data used for machine learning;
A learning support system comprising a control unit for generating a predictive model that predicts the progress of each process,
The control section,
Using the design information, sequentially arrange in a model space a three-dimensional model that combines element models of architectural elements used in each process of the construction work ,
In the model space , two-dimensional images are generated by photographing the three-dimensional model at a plurality of viewpoints set at positions where a person in charge may take photographs at the construction site of the construction work, and two-dimensional images are generated by photographing the three-dimensional model. Recorded in the information storage unit,
A learning support system characterized in that a prediction model for predicting the progress status of each process from images taken during the construction work is generated using teacher data recorded in the teacher information storage unit. The learning support system according to claim 1, wherein the control unit generates a prediction model that predicts architectural elements included in the two-dimensional image. 3. The learning support system according to claim 1, wherein the control unit performs rendering using a light source arranged in the model space. 4. The learning support system according to claim 3, wherein the control unit sets the light source under illumination conditions based on a photographing time. 5. The learning support system according to claim 4, wherein the photographing time sets a plurality of time periods in which the person in charge may take photographs. 6. The control unit generates the two-dimensional image by arranging a three-dimensional model of temporary storage materials in the model space for each teacher data. The learning support system described in . 7. The learning support system according to claim 6 , wherein the control unit specifies architectural elements used in steps before and after creating the teacher data, and uses them as the temporary storage materials. a design information storage unit that records design information of a structure configured by a three-dimensional element model including architectural elements associated with each process of construction work ;
a teacher information storage unit that records teacher data used for machine learning;
A method for providing learning support using a learning support system comprising a control unit for generating a predictive model that predicts the progress status of each process, the method comprising:
The control section,
Using the design information, sequentially arrange in a model space a three-dimensional model that combines element models of architectural elements used in each process of the construction work ,
In the model space , two-dimensional images are generated by photographing the three-dimensional model at a plurality of viewpoints set at positions where a person in charge may take photographs at the construction site of the construction work, and two-dimensional images are generated by photographing the three-dimensional model. Recorded in the information storage unit,
A learning support method comprising: generating a predictive model that predicts the progress of each process from images taken during the construction work using teacher data recorded in the teacher information storage unit. a design information storage unit that records design information of a structure configured by a three-dimensional element model including architectural elements associated with each process of construction work ;
a teacher information storage unit that records teacher data used for machine learning;
A program for providing learning support using a learning support system comprising a control unit for generating a predictive model that predicts the progress of each process,
The control section,
Using the design information, sequentially arrange in a model space a three-dimensional model that combines element models of architectural elements used in each process of the construction work ,
In the model space , two-dimensional images are generated by photographing the three-dimensional model at a plurality of viewpoints set at positions where a person in charge may take photographs at the construction site of the construction work, and two-dimensional images are generated by photographing the three-dimensional model. Recorded in the information storage unit,
A learning support program that functions as a means for generating a predictive model that predicts the progress of each process from images taken during the construction work using the teacher data recorded in the teacher information storage unit. .

Images