Information processing device, information processing method, program, and welding system

Abstract

To obtain a more accurate detection model.SOLUTION: An information processing device includes a generation unit and a learning unit. The generation unit uses one or more pieces of first teacher data containing one or more welding images obtained by imaging an object to be welded and true values of the positions of multiple feature points in the welding image to generate one or more pieces of second teacher data containing a converted image obtained by converting the welding images so as to change the relative positions of the multiple feature points and the true values of the positions after the change. The learning unit learns a detection model that receives an input of the welding image and outputs the positions of the feature points by using the first teacher data and the second teacher data.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] Embodiments of the present invention relate to an information processing apparatus, an information processing method, a program, and a welding system.

Background Art

[0002] There is known a technique for detecting feature points from an image (welding image) of a welding target and controlling welding using the positions of the detected feature points. In the detection of feature points from an image, for example, a detection model learned using teacher data including a learning image and true values (correct data) of the positions of the feature points in the image is used.

[0003] As a learning method for reducing the load for acquiring teacher data, for example, a method of using an image obtained during the operation of a welding system as teacher data can be considered.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, with the above learning method, for example, it may not be possible to obtain a detection model that improves the accuracy of detecting feature points in a situation (irregular situation) different from normal operation.

[0006] An object of the present invention is to provide an information processing apparatus, an information processing method, a program, and a welding system that can obtain a more accurate detection model for use in welding control.

Means for Solving the Problems

[0007] The information processing device of this embodiment comprises a generation unit and a learning unit. The generation unit generates one or more second training data sets, which include a transformed image obtained by transforming the welding image to change the relative positions of the multiple feature points, using one or more first training data sets that include one or more welding images of an object to be welded and the true values ​​of the positions of multiple feature points in the welding image, and the true values ​​of the positions after the transformation. The learning unit learns a detection model that takes a welding image as input and outputs the positions of feature points, using the first training data and the second training data. [Brief explanation of the drawing]

[0008] [Figure 1] A schematic diagram of a welding system equipped with an information processing device according to an embodiment. [Figure 2] Block diagram of an information processing device. [Figure 3] Flowchart of the learning process. [Figure 4] Flowchart of the welding control process. [Figure 5] A diagram illustrating an example of a feature point. [Figure 6] A diagram illustrating an example of a feature point. [Figure 7] A diagram illustrating an example of a feature point. [Figure 8] A diagram showing an example of an image where detection errors occurred. [Figure 9] A diagram showing an example of a generated transformed image. [Figure 10] A diagram showing an example of a welding image. [Figure 11] A diagram showing an example of a transformed image with an added region. [Figure 12] Hardware configuration diagram of an information processing device. [Modes for carrying out the invention]

[0009] A preferred embodiment of the information processing device according to this invention will be described in detail below with reference to the attached drawings.

[0010] One of the expectations for feature point detection is to guide the welded area to the correct position. Therefore, high detection accuracy is required in irregular situations where control is needed to correct the position.

[0011] However, as mentioned above, using images captured during welding system operation as training data may not improve the robustness of feature point detection in irregular situations that differ from normal operation.

[0012] On the other hand, for example, it is possible to obtain training data that anticipates irregular situations by using images taken under intentionally created irregular circumstances, but this increases the processing load. Therefore, from the perspective of processing load, it is desirable to train the detection model (feature point detector) based on images taken during the operation of the welding system.

[0013] Therefore, in this embodiment, images (a group of images) that can be used as training data are generated from a group of images obtained during normal operation by image transformation that reproduces irregular situations, and a detection model is trained using the training data, including the generated images. This makes it possible to obtain a detection model with higher accuracy for use in welding control. For example, it is possible to obtain a detection model that realizes robust feature point detection that can detect with higher accuracy even in irregular situations.

[0014] Figure 1 is a schematic diagram illustrating the configuration of a welding system equipped with an information processing device according to an embodiment. The welding system 1 comprises an information processing device 10, a welding device 20, a storage device 30, a PoE (Power over Ethernet®) hub 40, and a PLC (Programmable Logic Controller) 50. The storage device 30 may be an external device connected to the information processing device 10 or built into the information processing device 10, and is not limited to this embodiment.

[0015] The welding device 20 welds and integrates two or more members. The welding device 20 performs, for example, arc welding or laser welding. Specifically, arc welding includes, for example, Tungsten Inert Gas (TIG) welding, Metal Inert Gas (MIG) welding, Metal Active Gas (MAG) welding, or carbon dioxide arc welding. Here, mainly, an example in which the welding device 20 performs TIG welding will be described.

[0016] The welding device 20 includes, for example, a head 21, an arm 22, a wire 23, an imaging unit 24, a lighting unit 25, and a welding control unit 26. The welding control unit 26 includes a power supply unit 26a, a gas supply unit 26b, and a control unit 26c.

[0017] The head 21 is provided with a tungsten electrode 21a. The tip of the electrode 21a is exposed from the head 21. For example, the head 21 is attached to an articulated arm 22 including a plurality of links. Or the head 21 may be provided on a welding torch grasped by an operator.

[0018] The power supply unit 26a is electrically connected to the electrode 21a and the welding target S. A voltage is applied between the electrode 21a and the welding target S by the power supply unit 26a, and arc discharge occurs. One of the electrode 21a and the welding target S is set to a common potential (for example, ground potential), and the power supply unit 26a may control only the potential of the other of the electrode 21a and the welding target S. <S

[0019] The gas supply unit 26b is connected to the head 21. The gas supply unit 26b supplies an inert gas to the head 21. Or the gas supply unit 26b may supply a mixed gas of an inert gas and an active gas. The gas supplied to the head 21 is blown toward the welding target S from the tip of the head 21 where the electrode 21a is exposed.

[0020] The tip of the wire 23 is positioned in the space where an arc discharge is occurring. The arc discharge melts the tip of the wire 23, and it drips onto the object to be welded S. The object to be welded S is welded as the molten wire 23 solidifies. The wire 23 is fixed to, for example, the arm 22 and is automatically supplied as the melting progresses.

[0021] The imaging unit 24 photographs the area where welding is taking place during welding. The imaging unit 24 photographs the welding area and acquires a still image. Alternatively, the imaging unit 24 may record a video. The imaging unit 24 extracts a portion of the video to acquire a still image. The imaging unit 24 is, for example, a camera including a CCD image sensor or a CMOS image sensor.

[0022] The illumination unit 25 illuminates the welding area during welding so that the imaging unit 24 can obtain a clearer image. If an image usable for subsequent processing can be obtained without illuminating the welding area, the illumination unit 25 may not be necessary.

[0023] The control unit 26c controls the operation of each of the above-described components of the welding apparatus 20. For example, the control unit 26c drives the arm 22 and generates an arc discharge to weld the welding target S along a predetermined direction. The control unit 26c may also control the settings of the imaging unit 24, the illumination unit 25, and so on.

[0024] The PoE hub 40 is a hub that connects the imaging unit 24, the information processing device 10, and the PLC 50. The PLC 50 is connected to the welding control unit 26 (or control unit 26c) of the welding apparatus 20.

[0025] Images captured by the imaging unit 24 are transmitted to the information processing device 10, for example, via the PoE hub 40, and stored in the storage device 30 connected to the information processing device 10. For example, the captured images are stored in the storage device 30 in association with the welding conditions at the time of capture and the shooting conditions at the time of capture.

[0026] Welding conditions include, for example, applied voltage, gas flow rate, current value, wire feed rate, or welding speed. Imaging conditions include, for example, settings for the imaging unit 24 such as exposure time, aperture, or sensitivity (ISO). Imaging conditions may also include settings for the illumination unit 25. For example, when pulsed current is supplied to the illumination unit 25, imaging conditions further include pulse width, pulse frequency, duty cycle, or peak value. Here, when multiple items are listed connected by "or," it means that all of those items may be included, or only some of those items may be included.

[0027] Figure 2 is a block diagram showing an example of the configuration of the information processing device 10. As shown in Figure 2, the information processing device 10 includes an acquisition unit 101, a generation unit 102, a learning unit 103, an output control unit 104, and a storage unit 121.

[0028] The acquisition unit 101 acquires various types of information used by the information processing device 10. For example, the acquisition unit 101 acquires one or more welding images (which may be still images or videos) of the object to be welded, captured by the imaging unit 24 of the welding apparatus 20. The acquisition unit 101 also acquires the specification of ground truth data representing the true values ​​of the positions of multiple feature points within the welding image.

[0029] The ground truth data is used in conjunction with the welding images to train a detection model, which is one of the image recognition techniques (image recognition tasks) that detects the location of specific feature points in an image. Specifically, one or more training data sets, including the welding image and the true values ​​of the locations of multiple feature points within that welding image, are used to train the detection model.

[0030] Therefore, for example, the user specifies the positions of the feature points in the image, corresponding to the number of feature points to be detected. Any method can be used to specify the positions of the feature points, but for example, methods such as directly inputting coordinate values ​​on the image, or specifying the positions in the image using a mouse or the like with a graphical user interface provided separately for specifying positions in the image, can be applied. The acquisition unit 101 acquires ground truth data representing the positions of the feature points specified in this way.

[0031] The acquisition unit 101 can acquire (generate) training data that includes images acquired from the welding apparatus 20 and specified ground truth data (true values ​​of the positions of multiple feature points). If the training data is created by an external device other than the welding system 1, for example, the acquisition unit 101 may acquire the training data from the external device. In this case, the acquisition unit 101 does not need to acquire welding images from the welding apparatus 20, nor does it need to acquire the specified ground truth data.

[0032] The generation unit 102 uses the training data TA (first training data) obtained by the acquisition unit 101 to generate new training data TB (second training data) which includes a transformed image obtained by transforming the welding image contained in the training data TA. For example, the generation unit 102 generates one or more training data TB which include a transformed image obtained by transforming the welding image so as to change the relative positions of multiple feature points, and the true values ​​of the changed positions.

[0033] The learning unit 103 trains a detection model using training data TA and training data TB. The detection model is a model that detects the position of specific feature points in an input image, which is one of the image recognition techniques. The detection model may have any structure, but for example, it may be a model using DarkPose, which is one of the methods that uses a convolutional neural network. Furthermore, the learning method used by the learning unit 103 may be any method that is applicable to the detection model adopted.

[0034] The output control unit 104 controls the output of various types of information used by the information processing device 10. For example, the output control unit 104 outputs information (such as parameters) related to the learned detection model to the welding device 20. This allows the welding device 20 to use the learned detection model to control welding. For example, the control unit 26c inputs an image (welding image) captured by the imaging unit 24 to the detection model and controls welding using the positions of multiple feature points output by the detection model.

[0035] Each of the above components (acquisition unit 101, generation unit 102, learning unit 103, and output control unit 104) can be implemented by, for example, one or more processors. For example, each of the above components may be implemented by having a processor such as a CPU (Central Processing Unit) execute a program, i.e., by software. Each of the above components may also be implemented by a dedicated processor such as an IC (Integrated Circuit), i.e., by hardware. Each of the above components may also be implemented by using a combination of software and hardware. When multiple processors are used, each processor may implement one of the above components, or two or more of the above components.

[0036] The storage unit 121 stores various types of information used in the various processes of the information processing device 10. For example, the storage unit 121 stores images acquired by the acquisition unit 101 and the designation of correct answer data. The information processing device 10 may be configured to use a storage device 30 instead of the storage unit 121.

[0037] The memory unit 121 and the storage device 30 can be composed of any commonly used storage medium, such as flash memory, memory cards, RAM (Random Access Memory), HDD (Hard Disk Drive), and optical discs.

[0038] Next, the learning process of the detection model by the information processing device 10 will be described. Figure 3 is a flowchart showing an example of the learning process in this embodiment.

[0039] The acquisition unit 101 acquires one or more welding images from the welding apparatus 20 and also acquires the positions of correct feature points specified by the user or the like (step S101). As a result, the acquisition unit 101 can acquire (generate) one or more training data TAs, which include the welding images and the correct data representing the specified positions.

[0040] The generation unit 102 generates a transformed image by converting the acquired welding image, and generates one or more training data TBs including the transformed image and the true values ​​of the modified feature point positions (step S102).

[0041] The learning unit 103 trains a detection model using the training data TA and the training data TB (step S103), and then terminates the training process.

[0042] Next, the welding control process using the detection model by the welding apparatus 20 will be described. Figure 4 is a flowchart showing an example of the welding control process in this embodiment.

[0043] The imaging unit 24 acquires images of the welding area, for example, in a time series (step S201). The control unit 26c detects multiple feature points from each acquired image, for example, using a trained detection model (step S202). Based on the detected feature points, the control unit 26c controls the welding by the welding apparatus 20 to weld at the optimal location (step S203).

[0044] Any method of controlling welding based on feature points is acceptable. For example, we will explain using a case where feature points corresponding to the following object are used. (OBJ1) Tip of electrode 21a (hereinafter referred to as electrode tip) (OBJ2) The tip of wire 23 (hereinafter, wire tip) (OBJ3) Beveled wall (OBJ4) Melting pool contour

[0045] In this case, the control unit 26c drives the arm 22 by feedback control so that, for example, the electrode tip and wire tip are positioned at the center of the groove wall. The control unit 26c also controls the speed at which the electrode tip and wire tip move in a direction along the groove wall so that the molten pool contour is in contact with the groove wall.

[0046] The following will further explain the details of the processing performed by each of the above parts. First, we will describe an example of a feature point used in welding system 1. Figures 5 to 7 are diagrams illustrating an example of a feature point.

[0047] Figure 5 shows an example of a welding image in which objects 501, 502, 503, and 504, corresponding to the four objects (OBJ1) to (OBJ4) described above, are superimposed. Figure 6 shows an example of a welding image in which feature points 601 and 602, corresponding to object 501 (electrode tip) and object 502 (wire tip), are detected. Figure 7 shows an example of a welding image in which feature points 701a, 701b, 702a, and 702b, corresponding to object 503 (groove wall), and feature points 711a, 711b, 712a, and 722b, corresponding to object 504 (molten pool contour), are detected.

[0048] Thus, an object for which feature points are to be detected may be represented by one feature point or by multiple feature points. For example, as shown in Figure 6, the electrode tip and wire tip are each represented by one feature point. Also, as shown in Figure 7, the two groove walls (left and right groove walls) are each represented by two feature points, and the molten pool contour is represented by four feature points.

[0049] Since strong light is emitted during welding, two welding images, such as those shown in Figures 6 and 7, can be obtained by, for example, taking images in two modes with different exposure times. Alternatively, images can be taken with a single exposure time, and all feature points can be detected from a single welding image.

[0050] If the generation unit 102 does not generate a converted image, that is, if a detection model trained using only welding images as training data is used, the accuracy of feature point detection may decrease. Figure 8 shows three examples of images in which detection errors occurred. In Figure 8, black circles represent the true values ​​(ground truth data) of the feature point locations, and white circles represent the feature point locations predicted by the detection model.

[0051] To minimize detection errors as shown in Figure 8, in this embodiment, the generation unit 102 generates a transformed image by applying image transformation to the welding image and uses it as training data. That is, the generation unit 102 uses the training data TA obtained by the acquisition unit 101 to generate new training data TB that includes the transformed image obtained by transforming the welding image. By transforming the welding image in this way, an irregular situation is simulated. However, it is not necessary to reproduce the appearance of the image to a human, for example; it is sufficient to reproduce an irregular situation for the detection model.

[0052] For example, the welding apparatus 20 melts the base material (wire 23) with the heat from the arc discharge generated from the electrode 21a to perform welding. Therefore, the welding apparatus 20 (control unit 26c) controls, for example, the position of the electrode 21a and the position of the wire 23. Although it varies slightly depending on the welding method, the positional relationship between the metals to be joined, the electrode 21a, and the wire 23 is obtained from an image and controlled to the optimal position. Therefore, for example, a situation where the positions of the electrode 21a and the wire 23 are far apart can be said to be an irregular situation.

[0053] An example of a method for reproducing such irregular situations, namely, an example of a method for generating a converted image by the generation unit 102, is described below.

[0054] (Generation method 1) The generation unit 102 divides the welding image into two partial images by a line passing through the midpoints of multiple feature points, and generates a transformed image by relatively moving the two partial images so that they are shifted in the direction of the line. For example, the generation unit 102 divides the welding image vertically around the midpoint between the feature point corresponding to the electrode 21a (feature point 601 in Figure 6) and the feature point corresponding to the wire 23 (feature point 602 in Figure 6), and randomly shifts the upper and lower images horizontally.

[0055] This allows a welding image in which the electrode 21a and wire 23 were aligned horizontally (horizontally in the image) to be transformed into a transformed image representing an irregular situation where they are separated to the left and right. Thus, image transformation involves changing the position of the target feature point from the position in the original welding image to the expected destination position.

[0056] Figure 9 shows an example of a transformed image generated by generation method 1. Note that Figure 9 also shows an example of a transformed image obtained by translating a partial image and then rotating it. Thus, the generation unit 102 may perform image transformations that include rotation.

[0057] If the positions of two feature points are not aligned, for example, in the horizontal (or vertical) direction of the image, the welding image may be divided into two sub-images using lines that pass through the midpoints of the multiple feature points, or lines that are perpendicular to the line segments connecting the multiple feature points.

[0058] Movement may result in the conversion image containing areas not present in the original welding image. For such areas, for example, the average value of the pixel values ​​of the entire image or a fixed pixel value may be set. If the welding image before conversion is an image cropped from a larger image, the generation unit 102 may obtain information about the relevant area from the image before cropping and set it in the conversion image.

[0059] A combination of multiple feature points whose relative positions are changed is, for example, predetermined. The generation unit 102 may select a combination of predetermined combinations to change the relative positions from among the predetermined combinations according to a predetermined rule. The rule is, for example, a rule for random selection.

[0060] The relative positions may be changed among multiple groups, each containing one or more feature points. That is, the generation unit 102 may generate training data TB that includes a transformed image in which the relative positions among multiple groups have been changed.

[0061] The method for changing the relative position (the method of change) may also be specified in advance, or a method selected according to a rule from among multiple specifications may be chosen. The method of change may include, for example, some or all of the frequency, movement range (amount of movement), and movement direction. The frequency may represent, for example, the proportion of times an image transformation is performed in a learning process that is repeated multiple times.

[0062] The movement range is, for example, the range of the amount of movement of the position to be moved. The generation unit 102 may determine the amount of position movement from the movement range according to a predetermined rule. In this case, the rule may be, for example, a rule that determines it to follow a Gaussian distribution. The movement range (amount of movement) may be determined to be within the controllable degrees of freedom of the welding apparatus 20.

[0063] (Generation method 2) Image transformation by generation method 1 results in a transformation that appears discontinuous to the human eye. The generation unit 102 may use an image transformation that results in a continuous transformation. For example, the generation unit 102 considers an object containing multiple feature points as a non-rigid body and generates a transformed image by nonlinearly transforming the welding image so that the relative positions of the multiple feature points change according to the deformation of the non-rigid body. A nonlinear transformation is, for example, a transformation using the B-spline method.

[0064] (Generation method 3) The generation unit 102 may generate the transformed image using an image generator (such as a convolutional neural network model) that takes the image before modification and the positions of the feature points after modification as input and outputs a transformed image.

[0065] (Generation method 4) The generation unit 102 may generate a transformed image in which regions are added between multiple feature points. For example, welding may be performed in multiple stages because the metal is not sufficiently filled in a single welding pass. In such a case, for example, after the first welding pass is completed, a situation may arise where the distance between one of the groove walls and the molten pool contour is greater than the distance between the other groove wall and the molten pool contour.

[0066] Figure 10 shows an example of a welding image illustrating this situation. In the example in Figure 10, the two gray lines represent the groove walls. To the left of the groove wall on the right, there is an area 1001 where metal has not been filled because welding has not been performed.

[0067] If a detection model that has not been trained using training data including such welding images is used, feature points corresponding to the groove walls may not be detected correctly. Therefore, in generation method 4, a transformed image is generated that is the same as such a welding image. For example, the generation unit 102 generates a transformed image in which a region is added between the two feature points corresponding to the two groove walls, respectively.

[0068] Figure 11 shows an example of a transformed image with such a region 1101 added. The pixel values ​​of the added region may be fixed values, or they may be values ​​calculated from the pixel values ​​of the image, such as the average of the pixel values ​​of the entire image.

[0069] (Generation method 5) The generation unit 102 may generate a converted image according to a set of feature points specified by the user or the like, and a specified method for modifying their positions. These specifications are acquired, for example, by the acquisition unit 101 and output to the generation unit 102.

[0070] The method for specifying feature points and modification methods can be any method, but for example, a method in which the user specifies them on a screen displayed on a display device can be applied. For example, the user can specify a specific feature point, such as the tip of an electrode, on the screen using an input device such as a mouse, and further specify the direction and amount of movement. In the case of an object represented by two points, such as a groove wall, the user can specify the two points as a group, and further specify the direction and amount of movement of that group.

[0071] In this way, the information processing device of this embodiment generates training data that reproduces various situations through image transformation from data obtained under limited conditions (such as normal operation), and trains a detection model using the generated training data. This makes it possible to obtain a detection model with higher accuracy. For example, it becomes possible to detect the position of feature points more accurately even in situations other than those for which data was obtained.

[0072] Next, the hardware configuration of the information processing device according to the embodiment will be described using Figure 12. Figure 12 is an explanatory diagram showing an example of the hardware configuration of the information processing device according to the embodiment.

[0073] The information processing device according to this embodiment includes a control device such as a CPU 51, a storage device such as a ROM (Read Only Memory) 52 and RAM 53, a communication I / F 54 that connects to a network for communication, and a bus 61 that connects each part.

[0074] The program to be executed by the information processing device according to this embodiment is provided pre-installed in a ROM 52 or the like.

[0075] The program executed by the information processing device according to this embodiment may be configured to be provided as a computer program product by recording it in an installable or executable file format onto a computer-readable recording medium such as a CD-ROM (Compact Disk Read Only Memory), a flexible disk (FD), a CD-R (Compact Disk Recordable), or a DVD (Digital Versatile Disk).

[0076] Furthermore, the program executed by the information processing device according to the embodiment may be stored on a computer connected to a network such as the Internet and provided by downloading it via the network. Alternatively, the program executed by the information processing device according to the embodiment may be provided or distributed via a network such as the Internet.

[0077] A program executed by the information processing device according to this embodiment can cause a computer to function as one of the parts of the information processing device described above. This computer can read a program from a computer-readable storage medium onto its main memory and execute it using the CPU 51.

[0078] While several embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents. [Explanation of symbols]

[0079] 10 Information Processing Devices 20 Welding equipment 21 heads 21a electrode 22 Arms 23 wires 24 Imaging Department 25 Lighting Section 26 Welding Control Unit 26a Power supply section 26b Gas Supply Department 26c Control unit 30 Storage device 40 PoE Hubs 50 PLC 101 Acquisition Department 102 Generation part 103 Learning Department 104 Output Control Unit 121 Storage section

Claims

1. A generation unit generates one or more second training data sets, which include one or more welding images of the object to be welded, one or more first training data sets including the true values ​​of the positions of multiple feature points in the welding images, a transformed image obtained by transforming the welding image to change the relative positions of multiple feature points, and the true values ​​of the positions after the change. A learning unit learns a detection model that takes the welding image as input and outputs the position of the feature points using the first training data and the second training data, An information processing device equipped with the following features.

2. The generation unit generates the second training data, which includes the transformed image in which the relative positions between a plurality of groups, each containing one or more of the feature points, have been changed. The information processing apparatus according to claim 1.

3. The converted image is obtained by dividing the welding image into two partial images by a line passing through the midpoints of the multiple feature points, and then relatively moving the two partial images so that they are shifted in the direction of the line. The information processing apparatus according to claim 1.

4. The converted image is an image in which a region is added between a plurality of feature points. The information processing apparatus according to claim 1.

5. The transformed image is an image obtained by nonlinearly transforming the welding image such that the object containing the plurality of feature points is considered as a non-rigid body, and the relative positions of the plurality of feature points change according to the deformation of the non-rigid body. The information processing apparatus according to claim 1.

6. The system further includes an acquisition unit that acquires the specification of a plurality of feature points whose relative positions are changed, and the specification of a method for changing the positions. The generation unit modifies the relative positions of the specified plurality of feature points according to the specified modification method. The information processing apparatus according to claim 1.

7. An information processing method performed by an information processing device, A generation step of generating one or more second training data sets, which include one or more welding images of the object to be welded, one or more first training data sets including the true values ​​of the positions of multiple feature points in the welding images, a transformed image obtained by transforming the welding image to change the relative positions of multiple feature points, and the true values ​​of the changed positions, A learning step in which a detection model is trained to input the welding image and output the position of the feature points using the first training data and the second training data, Information processing methods including

8. On the computer, A generation step of generating one or more second training data sets, which include one or more welding images of the object to be welded, one or more first training data sets including the true values ​​of the positions of multiple feature points in the welding images, a transformed image obtained by transforming the welding image to change the relative positions of multiple feature points, and the true values ​​of the changed positions, A learning step in which a detection model is trained to input the welding image and output the position of the feature points using the first training data and the second training data, A program to execute.

9. A welding system including an information processing device and a welding device, The aforementioned information processing device is A generation unit generates one or more second training data sets, which include one or more welding images of the object to be welded, one or more first training data sets including the true values ​​of the positions of multiple feature points in the welding images, a transformed image obtained by transforming the welding image to change the relative positions of multiple feature points, and the true values ​​of the positions after the change. The system includes a learning unit that learns a detection model that takes the welding image as input and outputs the position of the feature points using the first training data and the second training data, The welding apparatus is The imaging unit captures the aforementioned welding image, The system includes a control unit that controls welding using the positions of a plurality of feature points output by the learned detection model, Welding system.

Images