Tire performance prediction device, performance prediction method, and learned model generation method

Abstract

To provide a performance prediction device of a tire capable of estimating performance data of the tire from image data of the tire acquired based on a predetermined image acquisition condition of the same perspective or projection.SOLUTION: A derivation unit configured to input a datum 21 including an image datum 21A of a target tire acquired based on an image acquiring condition, into a learned model 24B that has been machine-learned so as to output a performance datum 22 of the tire when the datum 21 including the image datum 21A of the tire acquired based on a predetermined identical perspective or projected image acquiring condition is inputted, and cause the learned model LA to output the performance datum 22 of the target tire.SELECTED DRAWING: Figure 8

Description

[Technical Field]

[0001] The present invention relates to a tire performance prediction device, a performance prediction method, and a method for generating a trained model. [Background technology]

[0002] Patent Document 1 listed below describes a tire image recognition method. This method involves inputting an image of the contact patch of a target tire to be recognized into a machine learning model that has learned the relationship between an image representing the tire contact patch and the contour of the tire contact patch, deriving an output, and estimating the contour of the contact patch of the target tire based on the derived output. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2024-048936 Summary of the Invention [Problem to be solved by the invention]

[0004] Generally, generating a machine learning model requires a large number of training data sets, and image data included in each training data set may be acquired under various image acquisition conditions, including perspective and projection. Because the differences in these image acquisition conditions are extracted as features for machine learning, there is room for further improvement in estimating tire performance data.

[0005] The present invention has been devised in view of the above-described circumstances, and has as its main object to provide a tire performance prediction device capable of estimating tire performance data from tire image data acquired under predetermined identical perspective or projection image acquisition conditions. [Means for solving the problem]

[0006] The present invention is a tire performance prediction device that includes a derivation unit that inputs data including image data of a target tire acquired based on predetermined image acquisition conditions into a trained model that has been machine-learned to output performance data of the tire when data including image data of the tire acquired based on the same predetermined perspective or projection image acquisition conditions is input, and outputs the performance data of the target tire. [Effects of the Invention]

[0007] By adopting the above-described configuration, the tire performance prediction device of the present invention is able to estimate tire performance data from tire image data acquired based on the same predetermined perspective or projection image acquisition conditions. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a block diagram illustrating an example of a tire performance prediction device and a trained model generation device. [Figure 2] 1 is a tire meridian cross-sectional view showing an example of a tire. [Figure 3] FIG. 2 is a development view of the tread portion of the tire. [Figure 4] 1A and 1B are diagrams for explaining an example of image acquisition conditions, in which FIG. 1A is a diagram for explaining image acquisition conditions for perspective projection, and FIG. 1B is a diagram for explaining image acquisition conditions for parallel projection. [Figure 5] 1 is a flowchart illustrating an example of a processing procedure for a method for generating a trained model. [Figure 6] 10 is a flowchart showing an example of a processing procedure of a preparation step. [Figure 7] FIG. 10 is a diagram illustrating an example of a training dataset. [Figure 8] FIG. 1 is a conceptual diagram illustrating an example of a machine learning model (trained model). [Figure 9] 1 is a flowchart showing an example of a processing procedure of a tire performance prediction method. [Figure 10] FIG. 4 is a side view showing an example of a decorative pattern on the sidewall portion. [Figure 11] FIG. 10 is a diagram illustrating an example of a training dataset according to another embodiment of the present invention. [Figure 12] FIG. 10 is a conceptual diagram showing an example of a machine learning model 24A (trained model 24B) according to another embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of the present invention will be described with reference to the drawings. It should be understood that the drawings include exaggerated representations and representations that differ from the dimensional ratios of actual structures in order to facilitate understanding of the contents of the invention. Furthermore, identical or common elements are designated by the same reference numerals throughout the embodiments, and redundant explanations will be omitted. Furthermore, the specific configurations shown in the embodiments and drawings are for the purpose of understanding the contents of the present invention, and the present invention is not limited to the specific configurations shown in the drawings.

[0010] [Tire performance prediction device] FIG. 1 is a block diagram showing an example of a tire performance prediction device 1A and a trained model generation device 1B. The tire performance prediction device (hereinafter, sometimes referred to as a "prediction device") 1A is a device for predicting the performance of a target tire (a tire to be evaluated). The prediction device 1A of this embodiment is also configured as a device (hereinafter, sometimes referred to as a "generation device") 1B for generating a trained model. Note that the prediction device 1A and the generation device 1B may be configured separately.

[0011] The prediction device 1A (generation device 1B) of this embodiment is configured as a general-purpose computer 1. Examples of the computer 1 include a desktop personal computer, a laptop personal computer, a tablet, a smartphone, and a cloud server.

[0012] The prediction device 1A (generation device 1B) of this embodiment includes an input device 3, an output device 4, and a calculation processing device 5.

[0013] [Input section, output section, arithmetic processing unit] The input device 3 is configured as an input device. Examples of this input device include a keyboard or a mouse. The output device 4 is configured as an output device. Examples of this output device include a display device or a printer. The arithmetic processing device 5 is used to predict the performance of a target tire. The arithmetic processing device 5 of this embodiment is configured to include one or more processors (arithmetic units) 5A that perform various calculations, a storage unit 5B in which data, programs, etc. are stored, and a working memory 5C.

[0014] [Processor] The processor 5A in this embodiment is configured as a central processing unit (CPU), but is not particularly limited thereto and may be configured as, for example, a microprocessor or other processing unit. In this embodiment, a mode is exemplified in which the tire performance prediction method (hereinafter sometimes referred to as the "prediction method") and the trained model generation method (hereinafter sometimes referred to as the "generation method") are executed by one processor 5A, but they may also be executed by multiple processors (parallel processing).

[0015] [Storage] The storage unit 5B is a non-volatile information storage device formed of, for example, a magnetic disk, an optical disk, an SSD, etc. The storage unit 5B is provided with a data unit 7 and a program unit 8.

[0016] [Data section] The data unit 7 of this embodiment is for storing data and the like necessary for executing the prediction method and the generation method. The data unit 7 of this embodiment includes a target data storage unit 7A, a target performance data storage unit 7B, a learning dataset storage unit 7C, and a model storage unit 7D. Note that the data unit 7 is not limited to this configuration, and may include storage units for storing other data as needed, or some of these may be omitted. The data stored in each storage unit 5B will be explained in each step of the prediction method and generation method described below.

[0017] [Program section] The program unit 8 of this embodiment is a program (application) required to execute the prediction method and the generation method. When such program unit 8 is executed by the processor (calculation unit) 5A, the computer 1 (prediction device 1A and generation device 1B) can function as a specific means. The program unit 8 of this embodiment includes an input unit 8A, a derivation unit 8B, a preparation unit 8C, a learning unit 8D, and an evaluation unit 8E. Note that the program unit 8 is not limited to this configuration, and other programs may be included as needed, or some of these may be omitted. The functions of each program unit 8 will be explained in the respective steps of the prediction method and generation method described below.

[0018] [tire] 2 is a tire meridian cross-sectional view showing an example of a tire 11. Although a pneumatic tire mounted on a passenger car is shown as an example of the tire 11 of this embodiment, the tire 11 is not limited to this. The tire 11 may be a tire of another category, such as a heavy-duty tire for trucks, buses, etc.

[0019] The tire 11 of this embodiment includes a tread portion 12, a pair of sidewall portions 13, and a pair of bead portions 14. The bead portions 14 have, for example, an annularly extending bead core 15. The tire 11 of this embodiment includes a carcass 16 and a belt layer 17.

[0020] [Carcass] The carcass 16 extends between the pair of bead portions 14. The carcass 16 includes at least one carcass ply 16A (one in this example).

[0021] The carcass ply 16A includes, for example, a main body portion 16a and a turned-up portion 16b. The main body portion 16a extends, for example, between the two bead portions 14. The turned-up portion 16b is continuous with the main body portion 16a and is turned up around the bead core 15 from the inner side to the outer side in the axial direction of the tire. A bead apex 18 extending from the bead core 15 to the outer side in the radial direction of the tire is disposed between the main body portion 16a and the turned-up portion 16b. The carcass ply 16A includes a plurality of carcass cords (not shown) arranged in parallel. These carcass cords may be made of organic fiber cords such as aramid or rayon.

[0022] [Belt layer] The belt layer 17 is disposed in the tread portion 12. The belt layer 17 includes at least one belt ply 17A, 17B, two in this embodiment. These belt plies 17A, 17B have a plurality of belt cords (not shown) arranged in parallel. Steel cords are used as the belt cords, but highly elastic organic fiber cords such as aramid and rayon can also be used as needed.

[0023] [Tread] Fig. 3 is a development view of the tread portion 12 of the tire 11. As shown in Fig. 2 and Fig. 3, the outer surface 12o of the tread portion 12 of this embodiment includes a contact patch 31 that contacts the road surface (not shown) and grooves 32 that are recessed radially inward from the contact patch 31. Furthermore, the outer surface 12o of the tread portion 12 may include sipes 33 that are narrower than the grooves 32 (for example, 1.5 mm or less).

[0024] As shown in FIG. 3 , the groove 32 of this embodiment includes a longitudinal groove 32A and a lateral groove 32B. The longitudinal groove 32A of this embodiment extends continuously in a zigzag pattern in the tire circumferential direction, but may extend continuously in a straight line. In this embodiment, four longitudinal grooves 32A are provided to divide the land portion 34 into five portions, but this is not limited to this. Furthermore, the lateral groove 32B extends in the land portion 34 in the tire axial direction. This divides the land portion 34 into a plurality of blocks 35.

[0025] In the tread portion 12 of this embodiment, a tread pattern is formed in which a plurality of pitches (not shown), which are pattern constituent units, are arranged in the tire circumferential direction by the longitudinal grooves 32A and the lateral grooves 32B. Each pitch may be of only one type having the same circumferential length, or may be of multiple types having different circumferential lengths.

[0026] [Sidewall] The sidewall portion 13 shown in Fig. 2 may be provided with, for example, a decorative pattern (not shown). Such a decorative pattern has the effect of, for example, utilizing the contrast of light, making molding defects such as bulges and dents that tend to occur on the surface of the sidewall portion 13 less noticeable. Such a decorative pattern improves the appearance (design) of the sidewall portion 13.

[0027] Furthermore, side blocks (not shown) may be formed on the sidewall portion 13. Such side blocks improve mud performance and performance on snow.

[0028] Incidentally, a large amount of training data sets is required to generate a machine learning model for predicting the performance, etc., of the tire 11. When these training data sets include, for example, image data (not shown) of the tire 11, the image data may be acquired under various image acquisition conditions including perspective projection and parallel projection (orthographic projection, axonometric projection, and oblique projection). The image acquisition conditions include various conditions such as perspective projection and parallel projection (orthogonal projection, axonometric projection, and oblique projection).

[0029] Fig. 4 is a diagram for explaining an example of "image acquisition conditions." Fig. 4(a) is a diagram for explaining perspective projection. Fig. 4(b) is a diagram for explaining parallel projection (orthogonal projection). These image acquisition conditions can be easily adjusted by editing image data using image editing software, etc.

[0030] In the perspective projection shown in FIG. 4(a), for example, distant parts (including the outer surface 12o of the tread portion 12 and the grooves 32) are displayed relatively small, thereby creating a sense of perspective. On the other hand, in the parallel projection (orthographic projection) shown in FIG. 4(b), the projection plane 37 and the projection line 38 are perpendicular, and parts are displayed at a constant size regardless of distance, so that parts can be displayed at the same actual ratio. By determining such image acquisition conditions in advance, the method of displaying images, including the ratio of the grooves 32 of the tire 11, can be standardized.

[0031] However, when image data is acquired under various image acquisition conditions, including perspective projection (Fig. 4(a)) and parallel projection (orthogonal projection) (Fig. 4(b)), the differences between these image acquisition conditions are extracted as machine learning features. These differences in image acquisition conditions are not correlated with tire performance data. For this reason, there is room for further improvement in the estimation of tire performance data.

[0032] In the prediction method (prediction device 1A) of this embodiment, data including image data of a target tire acquired under the same predetermined image acquisition conditions is input to a trained model that has been machine-learned in advance, thereby making it possible to estimate performance data of the target tire (hereinafter sometimes referred to as "performance data").

[0033] In this embodiment, the generation method is performed before the prediction method is performed to generate a trained model. Note that if a prediction model has already been generated when the prediction method is performed, the generation method may be omitted.

[0034] [How to generate a trained model] In the generation method of this embodiment, a trained model is generated. The trained model of this embodiment is a model that has been machine-learned so that, when data including image data of a target tire (a tire to be evaluated) acquired under the same predetermined image acquisition conditions is input, performance data of the target tire is output. To create such a trained model, it is effective to train the machine learning model using a training dataset that combines data and performance data acquired from multiple types of tires 11.

[0035] When data including image data of multiple tires 11 (two tires in this example) are input, the trained model of this embodiment outputs a relative performance value based on the performance of one of the multiple tires 11 as performance data. In this embodiment, the multiple tires 11 are exemplified as two tires, but this is not limited to this example and may be three or more tires. Furthermore, the trained model is not limited to inputting data including image data of multiple tires 11; for example, it may input data including image data of one tire 11 and output performance data for that tire.

[0036] 5 is a flowchart illustrating an example of a processing procedure of a method for generating a trained model. Each step of the generation method of this embodiment is executed by one or more processors 5A included in the generation device 1B (computer 1) illustrated in FIG.

[0037] [Prepare the training dataset (preparation process)] In the generation method of this embodiment, first, a training dataset is prepared (preparation step S1). In the preparation step S1 of this embodiment, first, a preparation unit 8C included in the program unit 8 shown in FIG. 1 is loaded into the working memory 5C. The preparation unit 8C is a program for preparing the training dataset. When this preparation unit 8C is executed by the processor 5A, the computer 1 (generation device 1B) can function as a means for preparing the training dataset.

[0038] As described above, the training data set is used for training the machine learning model and is a combination of data including tire image data acquired under the same predetermined image acquisition conditions and performance data of the tire 11.

[0039] The training dataset is preferably acquired based on multiple types of tires 11. The categories of the multiple types of tires 11 are not particularly limited. For example, from the viewpoint of improving the prediction accuracy of the trained model (machine learning model), the multiple types of tires 11 may be classified as the same category, or from the viewpoint of improving the versatility of the trained model (machine learning model), the multiple types of tires 11 may be classified as different categories. It is preferable that the multiple types of tires 11 differ from each other in at least one of, for example, category, tire size, tread pattern, and internal structure (including the carcass 16 and belt layer 17 shown in FIG. 2 ).

[0040] FIG. 6 is a flowchart showing an example of the processing procedure of the preparation step S1. FIG. 7 is a diagram showing an example of the training dataset 23. FIG. 7 shows one training dataset 23 as a representative example. As described above, when data 21 including image data 21A of two tires 11 is input, the trained model of this embodiment outputs a relative performance value 22A based on the performance of one tire 11 of the two tires 11 (hereinafter, sometimes referred to as the "reference tire 11s") as performance data 22. In order to generate such a trained model, the training dataset 23 of this embodiment includes, as data 21, image data 21A of two types of tires 11.

[0041] [Get data including tire image data] In the preparation step S1 of this embodiment, first, data 21 including image data 21A of the tire 11 acquired under the same predetermined image acquisition conditions is acquired (step S11). In step S11 of this embodiment, data 21 including image data 21A of each of a plurality of types of tires 11 is acquired.

[0042] The data 21 may be only tire image data (hereinafter sometimes referred to as "image data") 21A, or may further include other data in addition to the tire image data 21A. The data 21 of this embodiment includes only the image data 21A.

[0043] The image data 21A can be acquired as appropriate. For example, the image data 21A may be acquired by actually photographing the tire 11 shown in FIG. 2, or may be acquired from design data (CAD data) of the tire 11, or may be a hand-drawn image by an operator. The image data 21A may also be acquired from a tire model (not shown) modeled based on the design data of the tire 11. This tire model can be modeled as appropriate based on, for example, a known method (for example, the method described in Japanese Patent Application Laid-Open No. 2023-084539). The image data 21A of the present embodiment can be acquired from the tire model. This allows the image data 21A to be acquired at low cost and in a short time without actually manufacturing multiple types of tires 11.

[0044] The image data 21A is configured as, for example, a color image, a grayscale image, or a black-and-white image. The image data 21A of this embodiment is configured as a grayscale image. Such a grayscale image has the advantage of being able to identify the tire shape with high resolution compared to a black-and-white image. On the other hand, a grayscale image has the advantage of having a smaller data size compared to a color image.

[0045] The image data 21A can be acquired as appropriate. The image data 21A may be acquired of an uninflated tire 11 (shown in FIG. 2). This allows the image data 21A to be acquired in a shorter time than when the tire is inflated. The image data 21A may also be acquired of a tire 11 mounted on a rim and inflated to a predetermined internal pressure, or may be acquired of a tire 11 on which a predetermined load is applied. This stabilizes the shape of the tire 11, and further allows the image data 21A to be acquired under the same conditions for multiple types of tires 11. In this embodiment, the image data 21A is acquired based on a tire 11 (tire model) mounted on a rim and inflated to a predetermined internal pressure.

[0046] The rim (not shown) is exemplified as a genuine rim. A "genuine rim" is a rim defined for each tire by a standard system that includes the standard on which the tire 11 is based. Therefore, a genuine rim is, for example, a "standard rim" for JATMA, a "design rim" for TRA, or a "measuring rim" for ETRTO.

[0047] The internal pressure is exemplified as a normal internal pressure. The "normal internal pressure" is the air pressure determined for each tire by each standard in a standard system including the standard on which the tire 11 is based. Therefore, the normal internal pressure is, for example, the "maximum air pressure" in the case of JATMA, the maximum value listed in the table "TIRE LOAD LIMITS AT VARIOUS COLD INFLATION PRESSURES" in the case of TRA, and the "INFLATION PRESSURE" in the case of ETRTO.

[0048] The load is exemplified as a normal load. The "normal load" is the load determined for each tire by each standard in a standard system including the standard on which the tire 11 is based. Therefore, the normal load is, for example, "maximum load capacity" in the case of JATMA, the maximum value listed in the table "TIRE LOAD LIMITS AT VARIOUS COLD INFLATION PRESSURES" in the case of TRA, or "LOAD CAPACITY" in the case of ETRTO.

[0049] The image data 21A of this embodiment includes the contour of the tread portion 12. As a result, the image data 21A makes it possible to identify the pattern (tread pattern) engraved on the outer surface 12o of the tread portion 12 shown in Fig. 3. Such a pattern is constituted by, for example, grooves 32 (longitudinal grooves 32A and lateral grooves 32B) and sipes 33 provided on the outer surface 12o.

[0050] As in this embodiment, image data 21A including the contour of the tread portion 12 can identify contours including the contact patch 31, grooves 32 (including longitudinal grooves 32A and lateral grooves 32B), and sipes 33. These contours make it possible to identify the uneven shape (longitudinal grooves 32A, lateral grooves 32B, and sipes 33) of the tread portion 12. Such uneven shape affects the rigidity and contact shape of the tread portion 12, and therefore has a strong correlation with performance data 22 (in this example, performance values ​​of the mud performance of the tire) output from the trained model.

[0051] In the case where a tread pattern is formed by arranging a plurality of pitches (not shown) in the tire circumferential direction, as in the tread portion 12 of this embodiment, image data 21A for a specific pitch (one or more) may be acquired. Alternatively, image data 21A for all pitches may be acquired. In this embodiment, image data 21A for a plurality of pitches (e.g., three pitches) is acquired. In this case, the pitch for which image data 21A is acquired is appropriately selected according to performance data estimated by a trained model (machine learning model), and may include, for example, the pitch with the highest ground contact pressure among all pitches. This allows image data 21A to be acquired under the same conditions for a plurality of types of tires 11. Note that the pitch with the highest ground contact pressure can be easily identified, for example, by measuring or calculating the ground contact pressure of all pitches.

[0052] Image acquisition conditions for acquiring the image data 21A are determined as appropriate. When the outline of the tread portion 12 is included, as in the image data 21A of this embodiment, it is preferable to specify the same image acquisition conditions so that the tread ground contact edges 12t, 12t (shown in FIG. 3) on both axial sides of the tire are included. This makes it possible to reliably include the outline of the tread portion 12 in the image data 21A. The tread ground contact edges 12t, 12t correspond to the edges of the contact patch when a tire 11 in a normal state is loaded with 75% of the normal load and the tread portion 12 is placed in contact with a flat surface at a camber angle of 0°.

[0053] The image acquisition conditions in this embodiment are parallel projection (orthogonal projection) shown in Fig. 4(b) , which is preferable in that it can represent the grooves 32 that make up the tread portion 12 of the tire 11 at the same actual proportions.

[0054] As described above, in step S11 of this embodiment, image data 21A of multiple types of tires 11 are acquired based on the same predetermined image acquisition conditions (in this example, parallel projection (orthogonal projection)). This allows the ranges of multiple types of tires 11 to be unified in each image data 21A. By using a learning dataset 23 including such image data 21A for training a machine learning model, differences in image acquisition conditions between the image data 21A, 21A that are uncorrelated with the performance (performance data 22) of the tire 11, can be suppressed from being extracted as machine learning features.

[0055] In order to efficiently extract the feature amount of the image data 21A, the image data 21A may be acquired based on other predetermined conditions in addition to the image acquisition conditions (parallel projection (orthogonal projection) in this example) shown in Fig. 4. Such conditions preferably include at least one of a brightness condition, a size condition, and an image processing condition. Furthermore, the conditions may include a distance condition.

[0056] The brightness conditions include the illuminance of the subject 20. In this embodiment, the illuminance can be adjusted to a predetermined value, for example, by using a light (not shown) for photography or by using modeling software for the tire model. By predetermining such brightness conditions (illuminance), differences in brightness conditions between the image data 21A, 21A that are not correlated with tire performance can be suppressed from being extracted as machine learning features.

[0057] The size conditions include the size of the image data 21A (in this example, the number of vertical and horizontal pixels). In this embodiment, the size can be adjusted to a predetermined size by setting the camera, modeling software, etc. By predetermining such size conditions (sizes), differences in size conditions between the image data 21A, 21A that are not correlated with tire performance can be suppressed from being extracted as machine learning features.

[0058] The image processing conditions include, for example, image processing such as adjusting the color tone of the image data 21A and removing noise. In this embodiment, predetermined color tone and noise removal are possible by adjusting the editing process using, for example, commercially available image editing software. By predetermining such image processing conditions, differences in image processing conditions between the image data 21A, 21A that are not correlated with tire performance can be suppressed from being extracted as machine learning features.

[0059] The distance condition includes, for example, the shortest distance L1 from the projection plane 37 to the subject 20 (tire 11 in this example) shown in Fig. 4. By predetermining such shortest distance L1, it becomes possible to match the scale between the image data 21A, 21A. By predetermining such distance condition, it is possible to suppress the scale difference between the image data 21A, 21A that is not correlated with tire performance from being extracted as a feature for machine learning.

[0060] The image data 21A in this embodiment includes the contours of the tread portion 12 on both sides in the tire axial direction with respect to the tire equator C (shown in FIG. 3), but is not particularly limited thereto. The image data 21A may include, for example, the contour of the tread portion 12 on either side in the tire axial direction. Data 21 including the image data 21A for multiple types of tires 11 is input to the learning dataset storage unit 7C shown in FIG. 1.

[0061] [Get tire performance data] Next, in the preparation step S1 of this embodiment, performance data 22 (shown in FIG. 7) of the tire 11 is acquired (step S12). In step S12 of this embodiment, the performance data 22 is acquired for each of a plurality of types of tires 11.

[0062] The performance data 22 is not particularly limited as long as it relates to the performance of the tire 11. As shown in Fig. 7, when the image data 21A includes the outline of the tread portion 12, the performance data 22 preferably includes a performance value related to at least one of the drainage performance, mud performance, snow performance, and ice performance of the tire 11. These performance data 22 vary depending on the uneven shape (rigidity) of the tread portion 12, and therefore have a strong correlation with the image data 21A including the outline of the tread portion 12, which can identify the uneven shape. In addition, since the outline of the tread portion 12 has a strong correlation with the appearance of the tire 11, the performance data 22 may also include a rating related to the appearance of the tire 11.

[0063] The performance data 22 can be appropriately acquired by known procedures. The performance value of drainage performance can be acquired, for example, using a known inside-drum testing machine based on the running speed at which the tire 11 hydroplaned. The performance values ​​of mud performance, snow performance, and ice performance can be acquired, for example, by mounting the tire 11 on all wheels of a vehicle and conducting a sensory evaluation by a test driver. These performance values ​​may be, for example, relative evaluations based on predetermined standards or may be actual measured values. Furthermore, the performance data 22 can be acquired by simulation using a computer 1 (shown in FIG. 1). For such simulations, the above-described tire model is preferably used.

[0064] The performance data 22 may include a performance value relating to at least one of the drainage performance, mud performance, snow performance, and ice performance of the tire 11, or may include all of the performance values, or may further include other performance values. In this embodiment, the performance data 22 obtained is a performance value relating to mud performance.

[0065] The performance values ​​of drainage performance, mud performance, snow performance, and ice performance tend to be affected by road surface conditions and weather conditions. Therefore, it is preferable to acquire these performance values ​​based on at least one of predetermined road surface conditions and weather conditions. This allows performance data 22 to be acquired for multiple types of tires 11 under the same conditions (road surface conditions and weather conditions). Note that road surface conditions include, for example, the depth of a water film and a test course. Weather conditions include, for example, temperature, humidity, and solar radiation.

[0066] In step S12 of this embodiment, performance data 22 is acquired for each of a plurality of types of tires 11. The performance data 22 is input to the learning data set storage unit 7C shown in FIG.

[0067] [Create a training dataset] Next, in the preparation step S1 of this embodiment, as shown in FIG. 7, a learning dataset 23 is created by combining data 21 including image data 21A and tire performance data 22 (step S13).

[0068] As described above, when data 21 including image data 21A, 21A of two tires 11, 11 is input to the trained model of this embodiment, a relative performance value based on the performance of one of the tires 11 is output as performance data 22. To generate such a trained model, the data 21 of the training dataset 23 includes image data 21A, 21A of two tires 11, 11 selected from multiple types of tires 11. Furthermore, the performance data 22 of the training dataset 23 includes a relative performance value 22A (in this example, the relative performance value of the other tire 11) for the two selected tires 11, 11 based on the mud performance of one tire 11 (reference tire 11s).

[0069] The relative performance value 22A may be, for example, an index with the mud performance of one tire 11 (reference tire 11s) as a reference (e.g., "100"), or may be a pass / fail judgment result. In this embodiment, the relative performance value 22A is an index. In FIG. 7, the relative performance value (index) of the mud performance of the lower tire 11 with respect to the reference tire 11s displayed on the upper side is "50". This indicates that the lower tire 11 has lower mud performance than the upper tire 11 (reference tire 11s).

[0070] In this embodiment, one training data set 23 is created for two tires 11, 11 selected from multiple types of tires 11. Such two tires 11, 11 may be selected for all combinations of multiple types of tires 11, or may be selected for some combinations. In this embodiment, two tires 11, 11 are selected for all combinations of multiple types of tires 11. This allows multiple training data sets 23 to be created efficiently.

[0071] In this embodiment, in step S2, which will be described later, a plurality of training data sets 23 are used to train the machine learning model. As a result, when data 21 including image data 21A, 21A of two tires 11, 11 is input, it becomes possible to generate a trained model that can output performance data 22 of the tire 11 (in this example, a relative performance value 22A of mud performance). The plurality of training data sets 23 are input to the training data set storage unit 7C shown in FIG. 1.

[0072] [Training the machine learning model] Next, in the generation method of this embodiment, training of a machine learning model is performed using the training dataset 23 shown in Fig. 7 (step S2). In step S2, the machine learning model is trained so that when data 21 including image data 21A of a target tire (image data 21A, 21A of two target tires) is input, performance data 22 of the target tire (in this example, relative performance value 22A of mud performance) is output.

[0073] In step S2 of this embodiment, first, the training dataset 23 (shown in FIG. 7) stored in the training dataset storage unit 7C shown in FIG. 1 and a training unit 8D included in the program unit 8 are loaded into the working memory 5C. The training unit 8D is a program for training a machine learning model using the training dataset 23. Execution of this training unit 8D by the processor 5A causes the computer 1 (generation device 1B) to function as a means for training the machine learning model.

[0074] The machine learning model is not particularly limited as long as it is capable of performing machine learning such that, when data 21 including image data 21A of a target tire is input, performance data 22 of the target tire is output. In this embodiment, it is desirable to use a deep learning model equipped with a neural network as the machine learning model. This enables the machine learning model to extract the feature quantities of the input data 21 and estimate the performance data 22. Such a machine learning model can be appropriately configured based on a known method such as CNN.

[0075] In this embodiment, image data (in this example, a grayscale image) 21A is input to the machine learning model as data 21. In this case, the machine learning model (deep learning model) is preferably configured as a convolutional neural network (CNN) capable of extracting features of the image data 21A.

[0076] The convolutional neural network (CNN) may be a publicly known one, or a customized version of the publicly known one. The machine learning model (trained model) of this embodiment may be a customized version of VGG16, a publicly known CNN. Figure 8 is a conceptual diagram showing an example of a machine learning model 24A (trained model 24B).

[0077] The machine learning model 24A (trained model 24B) includes an input layer 25, an output layer 26, a convolution layer 27, a pooling layer 28, and a fully connected layer 29.

[0078] The input layer 25 can receive image data 21A, data 21 including 21A. Unlike VGG16, the output layer 26 (third fully connected layer 29C) has one node. As a result, the machine learning model 24A of this embodiment can output one piece of performance data 22 (in this example, a relative performance value 22A of mud tire performance) by receiving data 21 including image data 21A, 21A of two tires as input to the input layer 25.

[0079] In the convolutional layer 27, filtering is performed on the two image data 21A and the data 21 including 21A. As a result, a feature map is generated. A convolutional layer 27 is defined in each of the multiple convolutional blocks 30. These convolutional blocks 30 include a first convolutional block 30A, a second convolutional block 30B, a third convolutional block 30C, a fourth convolutional block 30D, and a fifth convolutional block 30E. Two convolutional layers 27 are defined in each of the first convolutional block 30A and the second convolutional block 30B. Three convolutional layers 27 are defined in each of the third convolutional block 30C, the fourth convolutional block 30D, and the fifth convolutional block 30E.

[0080] The pooling layer 28 reduces the size of the feature map generated in the convolutional layer 27. The pooling layer 28 includes a first pooling layer 28A, a second pooling layer 28B, a third pooling layer 28C, a fourth pooling layer 28D, and a fifth pooling layer 28E. The first pooling layer 28A is defined between the first convolutional block 30A and the second convolutional block 30B. The second pooling layer 28B is defined between the second convolutional block 30B and the third convolutional block 30C. The third pooling layer 28C is defined between the third convolutional block 30C and the fourth convolutional block 30D. The fourth pooling layer 28D is defined between the fourth convolutional block 30D and the fifth convolutional block 30E. The fifth pooling layer 28E is defined between the fifth convolutional block 30E and the fully connected layer 29.

[0081] The fully connected layer 29 connects the feature maps and outputs performance data 22 (in this example, relative performance values ​​22A of mud performance) to the output layer 26. The fully connected layer 29 of this embodiment includes a first fully connected layer 29A, a second fully connected layer 29B, and a third fully connected layer 29C. Of the first to third fully connected layers 29A to 29C, the third fully connected layer 29C is configured as the output layer 26.

[0082] In step S2 of this embodiment, first, one training data set 23 shown in Fig. 7 is selected from a plurality of training data sets 23 (not shown). Then, data 21 (in this example, image data 21A, 21A of two tires 11, 11) included in the selected training data set 23 is input to the machine learning model 24A shown in Fig. 8. As a result, performance data 22 (in this example, relative performance value 22A of mud performance) estimated by the machine learning model 24A can be output.

[0083] Next, in step S2 of this embodiment, the error between the output performance data (estimated data) 22 and the performance data (ground truth data) 22 included in the training dataset 23 shown in Figure 7 is calculated. Then, various parameters (e.g., weighting coefficients, biases, etc.) of the machine learning model 24A are updated so as to minimize this error. In this way, the series of steps from inputting the training dataset 23 to updating the parameters is performed for each of the multiple training datasets 23, thereby optimizing the machine learning model 24A.

[0084] In step S2 of this embodiment, a machine learning model 24A is optimized using multiple training datasets 23 (not shown). As a result, when data 21 including image data 21A of a target tire (in this example, image data 21A, 21A of two target tires) is input, a trained model 24B is generated that can output performance data 22 of the target tire (in this example, relative performance value 22A of mud performance). Such trained model 24B makes it possible to estimate the performance data 22 of the target tire without the need for the experience or intuition of an expert. The trained model 24B is input to the model storage unit 7D shown in FIG. 1.

[0085] As described above, the performance of the tire 11 shown in Fig. 2 (in this example, mud performance) affects the rigidity and contact shape of the tread portion 12. For this reason, the performance data 22 shown in Fig. 7 has a strong correlation with the image data 21A, 21A including the contour of the tread portion 12. By inputting data 21 including such image data 21A, 21A into the machine learning model 24A (trained model 24B) shown in Fig. 8, feature quantities for estimating the performance data 22 can be effectively extracted.

[0086] Furthermore, the image data 21A, 21A shown in FIG. 7 are acquired based on the same predetermined image acquisition conditions (in this example, parallel projection (orthogonal projection) shown in FIG. 4(b)). Therefore, the image representation method, including the ratio of the tires 11, 11, can be unified in the image data 21A, 21A shown in FIG. 7. By using a training dataset 23 including these image data 21A, 21A for training a machine learning model 24A (shown in FIG. 8), differences in image acquisition conditions (representation method) that are uncorrelated with tire performance can be suppressed from being extracted as machine learning features. Therefore, the generation method (generation device 1B) of this embodiment makes it possible to generate a trained model 24B (shown in FIG. 7) that can accurately estimate performance data 22.

[0087] In this embodiment, the image data 21A, 21A are acquired based on predetermined brightness conditions, size conditions, and image processing conditions in addition to the image acquisition conditions shown in FIG. 4(b). This can prevent differences in brightness conditions, size conditions, and image processing conditions between the image data 21A, 21A from being extracted as machine learning features. Furthermore, in this embodiment, the image data 21A, 21A are acquired based on predetermined distance conditions (shortest distance L1). This can prevent differences in scale between the image data 21A, 21A from being extracted as machine learning features.

[0088] In this embodiment, the trained model 24B is generated based on a conventional model (VGG16), which allows the trained model 24B to be created in a short time.

[0089] The trained model 24B may be capable of outputting contribution data (not shown) that visualizes features that contributed to the output of the performance data 22 in the image data 21A, 21A of the target tire. Such contribution data is preferably output as a heat map of the portion of the image data 21A, 21A to which the performance data 22 contributed. Such visualization of features can be achieved, for example, by implementing the well-known Grad-CAM in the trained model 24B.

[0090] [Tire performance prediction method (first embodiment)] Next, an example of a processing procedure for a tire performance prediction method will be described. The prediction method of this embodiment uses a trained model 24B (shown in FIG. 7) generated based on the generation method shown in FIGS. 5 and 6.

[0091] 9 is a flowchart showing an example of a processing procedure of a tire performance prediction method. Each step of the prediction method of this embodiment is executed by one or more processors 5A included in the prediction device 1A (computer 1) shown in FIG.

[0092] [Enter data including an image showing the tire frame shape of the target tire] In the prediction method of this embodiment, first, data 21 including image data 21A of a target tire is input to the trained model 24B shown in Fig. 8 (step S3). The image data 21A is acquired based on the same predetermined image acquisition conditions shown in Fig. 4 (in this example, parallel projection (orthogonal projection) shown in Fig. 4(b)).

[0093] In step S3 of this embodiment, first, the trained model 24B (shown in FIG. 7) input to the model storage unit 7D shown in FIG. 1 and the input unit 8A included in the program unit 8 are loaded into the working memory 5C. The input unit 8A is a program for inputting data 21 including image data 21A of the target tire to the trained model 24B. When the input unit 8A is executed by the processor 5A, the computer 1 (prediction device 1A) can function as a means for inputting the data 21 to the trained model 24B.

[0094] The trained model 24B has been machine-trained to output performance data 22 (in this example, a relative performance value 22A of mud performance) of the tire 11 when data 21 including image data 21A of the tire 11 (image data 21A, 21A of two tires 11, 11) is input. Therefore, in step S3 of this embodiment, first, data 21 including image data 21A, 21A of the two target tires is acquired.

[0095] In step S3 of this embodiment, image data 21A, 21A of the two target tires shown in FIG. 8 are acquired based on the same predetermined image acquisition conditions (in this example, parallel projection (orthogonal projection) shown in FIG. 4(b)) in accordance with the processing procedure in step S11 of the preparation step S1 shown in FIG. 6. This allows the image representation method, including the ratio of the two target tires, to be unified in the image data 21A, 21A. Furthermore, by acquiring the image data 21A, 21A of the two target tires based on predetermined brightness conditions, size conditions, image processing conditions, and distance conditions, the illuminance, size, color, noise removal, and shortest distance L1 of the two target tires can be unified. The acquired image data 21A, 21A are input to the target data storage unit 7A (shown in FIG. 1).

[0096] Next, in step S3 of this embodiment, the image data 21A of the two target tires, data 21 including 21A, are input to the trained model 24B. This allows the trained model 24B to output performance data 22 of the target tires (in this example, a relative performance value 22A based on the mud performance of one target tire (reference tire 11s)).

[0097] [Output performance data of target tire] Next, in the prediction method of this embodiment, the performance data 22 of the target tire is output from the trained model 24B (step S4).

[0098] In step S4 of this embodiment, first, the derivation unit 8B included in the program unit 8 shown in Fig. 1 is loaded into the working memory 5C. The derivation unit 8B is a program for outputting performance data of the target tire from the trained model 24B. Execution of this derivation unit 8B by the processor 5A causes the computer 1 to function as a means for outputting performance data 22 (shown in Fig. 8) of the target tire.

[0099] In this embodiment, in step S3, as shown in Fig. 8, data 21 including image data 21A, 21A of two target tires is input to a trained model 24B. As a result, in step S4, a relative performance value 22A (in this example, the relative performance value of the other target tire) for the two target tires can be output from the trained model 24B, using the mud performance of one target tire (reference tire 11s) as a reference. Therefore, the prediction method of this embodiment makes it possible to estimate and compare performance data 22 for two target tires without requiring the experience or intuition of a skilled person.

[0100] In this embodiment, image data 21A, 21A of the two target tires shown in FIG. 8 are acquired based on the same predetermined image acquisition conditions (in this example, parallel projection (orthogonal projection) shown in FIG. 4(b)). This allows the method of representing the images of the two target tires to be unified. This prevents differences in the image acquisition conditions from being extracted as feature amounts in the image data 21A, 21A of the two target tires, allowing the performance data 22 to be estimated with high accuracy.

[0101] Furthermore, in this embodiment, the image data 21A, 21A of the two target tires are acquired based on predetermined brightness conditions, size conditions, image processing conditions, and distance conditions. This allows the illuminance, size, color, noise removal, and shortest distance L1 of the two target tires to be unified. Therefore, differences in brightness conditions, size conditions, image processing conditions, and distance conditions (scale) between the image data 21A, 21A of the two target tires are extracted as feature amounts but are suppressed, allowing the performance data 22 to be estimated with high accuracy.

[0102] The performance data 22 (in this example, the relative performance value 22A of mud performance) output from the trained model 24B is stored in the target performance data storage unit 7B shown in Fig. 1. In addition, in step S4, the performance data 22 output from the trained model 24B may be output (displayed) to an output device 4 including, for example, a display device. This allows an operator or the like to understand the performance data 22 and further enables evaluation of the performance of the target tire.

[0103] Furthermore, in step S4, the derivation unit 8B shown in FIG. 1 (trained model 24B shown in FIG. 8) may output contribution data (not shown) that visualizes the features that contributed to the output of the performance data. Such contribution data is output as a heat map of the portions of the image data 21A, 21A of the two target tires that the performance data 22 contributed to. Therefore, the portions that have a large impact on the performance of the tire 11 (in this example, mud performance) can be easily and visually grasped, which is useful for, for example, changing the design of the target tire.

[0104] [Manufacture target tires with good performance data] Next, in the prediction method of this embodiment, of the two target tires, the target tire with better performance data is manufactured (step S5).

[0105] In this embodiment, a relative performance value 22A (in this example, the relative performance value of the other target tire) based on the mud performance of one target tire (reference tire 11s) is output as the performance data 22. Based on this relative performance value, the target tire with better mud performance is selected from the two target tires. The target tire is then manufactured based on image data 21A of the selected target tire. In this way, a tire 11 having desired performance (for example, mud performance) can be designed and manufactured.

[0106] [Method for generating trained model (second embodiment)] In the above-described embodiments, as shown in FIG. 7, the image data 21A includes the outline of the tread portion 12, but the present invention is not limited to this. The image data 21A may also include the outline of the sidewall portion 13. With this image data 21A, it is possible to identify the decorative pattern (not shown) provided on the sidewall portion 13 shown in FIG. 2. FIG. 10 is a side view showing an example of the decorative pattern 36 on the sidewall portion 13.

[0107] As described above, the decorative pattern 36 has the effect of making less noticeable molding defects that tend to occur on the surface of the sidewall portion 13. Therefore, when the performance data 22 output from the trained model 24B shown in FIG. 8 includes a rating score regarding the tire appearance, the decorative pattern 36 has a strong correlation with the performance data 22.

[0108] [Preparing a learning dataset (preparation step) (second embodiment)] In the generation method of this embodiment, as in the previous embodiments, first, a training data set 23 is prepared (preparation step S1). As in the previous embodiments, this training data set 23 is preferably obtained based on multiple types of tires 11.

[0109] FIG. 11 is a diagram showing an example of a training dataset 23 according to another embodiment of the present invention. FIG. 11 shows one training dataset 23 as a representative example. As described above, when data 21 including image data 21A of two tires 11 is input, the trained model 24B according to this embodiment outputs a relative performance value 22A based on the performance of one of the two tires 11 (hereinafter, sometimes referred to as the "reference tire 11s") as performance data 22. In order to generate such a trained model, the training dataset 23 according to this embodiment includes, as data 21, image data 21A of two types of tires 11.

[0110] [Acquiring data including tire image data (second embodiment)] In the preparation step S1 of this embodiment, similar to the previous embodiments, data 21 including image data 21A of the tire acquired under the same predetermined image acquisition conditions (in this example, parallel projection (orthogonal projection) shown in FIG. 4(b)) is acquired (step S11). In step S11 of this embodiment, data 21 including image data 21A including the contour of the sidewall portion 13 is acquired. In this embodiment, similar to the previous embodiments, data 21 including image data 21A of multiple types of tires 11 is acquired.

[0111] The image data 21A may be acquired by actually photographing the sidewall portion 13 of the tire 11, may be acquired from design data (CAD data) of the tire 11, or may be a hand-drawn image by an operator. The image data 21A may also be acquired from the sidewall portion of a tire model (not shown) modeled based on the design data of the tire 11. The image data 21A is configured as a grayscale image, but may also be configured as a color image or a black and white image.

[0112] The image data 21A is acquired as appropriate. The image data 21A may be acquired of an unempuffed tire 11 (shown in FIG. 2). The image data 21A may also be acquired of a tire 11 mounted on a rim and inflated to a predetermined internal pressure, or may be acquired of a tire 11 on which a predetermined load is applied. This stabilizes the shape of the tire 11, and further allows the image data 21A to be acquired under the same conditions for multiple types of tires 11. In this embodiment, the image data 21A including the contour of the sidewall portion 13 is acquired based on a tire 11 (tire model) mounted on a rim and inflated to a predetermined internal pressure.

[0113] Image acquisition conditions for acquiring the image data 21A are determined as appropriate. When the image data 21A includes the contours of the sidewall portions 13, the image acquisition conditions (in this example, parallel projection (orthogonal projection) shown in FIG. 4(b)) may be specified so as to include, for example, the range from the outermost end of the tread portion 12 in the tire radial direction to the innermost end of the bead portions 14 in the tire radial direction. This makes it possible to reliably include the contours of the sidewall portions 13 in the image data 21A.

[0114] In step S11 of this embodiment, image data 21A of multiple types of tires 11 is acquired based on the same predetermined image acquisition conditions (in this example, parallel projection (orthogonal projection) shown in FIG. 4(b)), which can unify the method of representing images of the sidewall portions 13 of multiple types of tires 11. By using a training dataset 23 including such image data 21A for training a machine learning model, differences in image acquisition conditions between the image data 21A that are not correlated with tire performance can be suppressed from being extracted as machine learning features.

[0115] To efficiently extract the feature quantities of the image data 21A, the image data 21A may be acquired based on other predetermined conditions (for example, brightness conditions, size conditions, image processing conditions, and distance conditions (shortest distance L1 shown in FIG. 4)) in addition to the image acquisition conditions. This can prevent differences in other conditions between the image data 21A that are not correlated with tire performance from being extracted as machine learning features.

[0116] The image data 21A in this embodiment includes the contour of the sidewall portion 13 on one side in the tire axial direction, but is not particularly limited to this. The image data 21A may include, for example, the contours of the sidewall portion 13 on both sides in the tire axial direction. Data 21 including the image data 21A is input to the learning dataset storage unit 7C shown in FIG. 1.

[0117] [Acquiring tire performance data (second embodiment)] Next, in the generation method of this embodiment, similarly to the previous embodiments, performance data 22 (shown in FIG. 6) of each tire 11 is acquired (step S12). In step S12 of this embodiment, performance data 22 is acquired for each of a plurality of types of tires 11.

[0118] When the image data 21A includes the outline of the sidewall portion 13 as in this embodiment, it is preferable that the performance data 22 include a rating for the appearance of the tire 11. The appearance of such a tire 11 changes depending on the decorative pattern 36 provided on the sidewall portion 13, and therefore there is a strong correlation with the image data 21A including the outline of the sidewall portion 13 from which the decorative pattern 36 can be identified. Furthermore, since the outline of the sidewall portion 13 makes it possible to identify side blocks (not shown), the performance data 22 may include mud performance and snow performance.

[0119] The rating on the appearance of the tire 11 can be obtained as needed. The rating is preferably given by a predetermined evaluator. This can prevent bias in the ratings, which tends to occur when ratings are given by an unspecified number of evaluators. It is preferable that the evaluators are selected so that the ages and genders of the evaluators are balanced. This can effectively prevent bias in the ratings.

[0120] The ratings can be acquired as appropriate. The ratings are preferably assigned based on an evaluation axis including at least one of "preference," "cool," and "cute." In this embodiment, the evaluation axis is exemplified as "preference." The ratings may also be assigned based on, for example, any one of a 5-point system, a 10-point system, and a 100-point system. The ratings assigned by each evaluator for multiple types of tires are then averaged. This allows performance data 22 including the ratings to be acquired for each of the multiple types of tires 11. The performance data 22 is input to the learning dataset storage unit 7C shown in FIG. 1.

[0121] [Creating a training dataset (second embodiment)] Next, in the preparation step S1 of this embodiment, as shown in FIG. 11, a learning dataset 23 is created by combining data 21 including image data 21A and tire performance data 22 (step S13).

[0122] The data 21 of the training dataset 23 in this embodiment includes image data 21A, 21A of two tires 11, 11 selected from multiple types of tires 11. Furthermore, the performance data 22 of the training dataset 23 includes a relative performance value 22A for these tires 11, 11, based on the rating of one tire 11 (reference tire 11s) (in this example, the relative performance value of the other tire 11). The relative performance value 22A may be, for example, an index based on the rating of the appearance of one tire 11 (reference tire 11s) (for example, "100"), or may be a pass / fail judgment result. In this embodiment, the relative performance value 22A is exemplified as an index.

[0123] 11, the relative performance value (index) of the rating score for the lower tire 11 relative to the reference tire 11s displayed on the upper side is "103." This indicates that the lower tire 11 has a higher rating score for the appearance than the upper tire 11 (reference tire 11s).

[0124] In this embodiment, similarly to the previous embodiments, two tires 11, 11 are selected from all combinations of multiple types of tires 11. In this way, multiple learning datasets 23 can be created.

[0125] In this embodiment, in step S2 described below, a plurality of training data sets 23 are used to train the machine learning model. As a result, when data 21 including image data 21A, 21A of two tires 11, 11 is input, it becomes possible to generate a trained model that can output performance data 22 (relative performance value 22A of appearance-related ratings) of the tire 11. The plurality of training data sets 23 are input to the training data set storage unit 7C shown in FIG. 1.

[0126] [Implementing machine learning model learning (second embodiment)] Next, in the generation method of this embodiment, learning of a machine learning model 24A is performed (step S2) using the learning dataset 23 shown in Figure 11. Figure 12 is a conceptual diagram showing an example of a machine learning model 24A (trained model 24B) according to another embodiment of the present invention.

[0127] In step S2 of this embodiment, similarly to the previous embodiments, a machine learning model 24A is optimized using a plurality of training data sets 23 (not shown). As a result, when data 21 including image data 21A of a target tire (image data 21A, 21A of two target tires) is input, a trained model 24B capable of outputting performance data 22 of the target tire (relative performance value 22A of appearance-related ratings) can be generated. The trained model 24B is input to the model storage unit 7D shown in FIG. 1.

[0128] As described above, the performance of the tire 11 shown in FIG. 10 (in this example, the appearance of the tire 11) changes depending on the decorative pattern 36 provided on the sidewall portion 13. For this reason, the performance data 22 shown in FIG. 11 has a strong correlation with the image data 21A, 21A including the contour of the sidewall portion 13. By inputting the data 21 including such image data 21A, 21A into the machine learning model 24A (trained model 24B) shown in FIG. 12, feature quantities for estimating the performance data 22 can be effectively extracted.

[0129] Furthermore, since the image data 21A, 21A are acquired based on the same predetermined image acquisition conditions (in this example, parallel projection (orthogonal projection) shown in FIG. 4(b)), the range in which the tires 11, 11 are captured can be unified. When the training dataset 23 including such image data 21A, 21A is used to train the machine learning model 24A, differences in image acquisition conditions between the image data 21A, 21A that are uncorrelated with tire performance can be suppressed from being extracted as machine learning features. Furthermore, the image data 21A, 21A are acquired based on other predetermined conditions (in this example, brightness conditions, size conditions, image processing conditions, and distance conditions) in addition to the image acquisition conditions. This can suppress differences in other conditions between the image data 21A, 21A from being extracted as machine learning features. Therefore, the generation method (generation device 1B) of this embodiment makes it possible to generate a trained model 24B that can accurately estimate the performance data 22.

[0130] [Tire performance prediction method (second embodiment)] In the prediction method of this embodiment, as in the previous embodiments, a trained model 24B (shown in FIG. 12) generated based on the generation method shown in FIG. 5 is used.

[0131] In the prediction method of this embodiment, as in the previous embodiments, in step S3 shown in Fig. 8, data 21 including image data 21A, 21A of tires 11 (in this example, two tires 11, 11) is input to the trained model 24B shown in Fig. 12. This makes it possible to output performance data 22 of the target tire (in this example, relative performance value 22A of the appearance-related rating) from the trained model 24B.

[0132] Next, in the prediction method of this embodiment, as in the previous embodiments, in step S4 shown in Fig. 8, performance data 22 of the target tire (in this example, relative performance value 22A of the appearance-related rating) is output from trained model 24B shown in Fig. 12. As a result, the prediction method of this embodiment makes it possible to estimate performance data 22 of the target tire without requiring the experience or intuition of a skilled person.

[0133] In step S4, the derivation unit 8B (trained model 24B) may output contribution data (not shown) that visualizes the features that contributed to the output of the performance data. Such contribution data makes it easy to visually grasp the parts of the tire 11 that have a large impact on its appearance (in this example, personal preference), which is useful for, for example, changing the design of the decorative pattern 36 (shown in FIG. 10) of the target tire.

[0134] Then, of the two target tires, the target tire with better performance data is manufactured (step S5). In this way, a tire 11 having desired performance (for example, good appearance) can be designed and manufactured.

[0135] Although a particularly preferred embodiment of the present invention has been described in detail above, the present invention is not limited to the illustrated embodiment and can be modified and implemented in various ways.

[0136] [Note] The present invention includes the following aspects.

[0137] [Invention 1] a derivation unit that inputs data including image data of a target tire acquired based on the image acquisition conditions into a trained model that has been machine-learned so as to output performance data of the tire when data including image data of the tire acquired based on the same predetermined perspective or projection image acquisition conditions is input, and outputs the performance data of the target tire. Tire performance prediction device. [Invention 2] The tire performance prediction device according to the first aspect of the present invention, wherein the image data is acquired based on at least one of predetermined brightness conditions, size conditions, and image processing conditions in addition to the image acquisition conditions. [Invention 3] the data includes image data of a plurality of tires acquired based on the image acquisition conditions, 3. The tire performance prediction device according to claim 1 or 2, wherein the performance data includes a relative performance value based on the performance of one tire of the plurality of tires. [Invention 4] 4. The tire performance prediction device according to any one of claims 1 to 3, wherein the image data includes a contour of a tread portion. [Invention 5] 5. The tire performance prediction device according to any one of claims 1 to 4, wherein the image data includes a profile of a sidewall portion. [Invention 6] 6. The tire performance prediction device according to any one of claims 1 to 5, wherein the performance data includes a performance value relating to at least one of drainage performance, mud performance, snow performance, and ice performance of the tire. [Invention 7] 7. The tire performance prediction device according to claim 6, wherein the performance value is obtained based on at least one of predetermined road surface conditions and weather conditions. [Invention 8] 8. The tire performance prediction device according to any one of claims 1 to 7, wherein the performance data includes a rating on the appearance of the tire. [Invention 9] The tire performance prediction device according to claim 8, wherein the score is given by a predetermined evaluator. [Invention 10] 10. The tire performance prediction device according to any one of claims 1 to 9, wherein the derivation unit further outputs contribution data that visualizes features that contributed to the output of the performance data. [Invention 11] 1. A method for predicting tire performance executed by one or more processors, comprising: a step of inputting data including image data of a target tire acquired based on the same predetermined perspective or projection image acquisition conditions into a trained model that has been machine-learned so as to output performance data of the tire when data including image data of the tire acquired based on the same predetermined perspective or projection image acquisition conditions is input; and outputting performance data of the target tire from the trained model. A method for predicting tire performance. [Invention 12] preparing a learning dataset that combines data including tire image data acquired under predetermined image acquisition conditions of the same perspective or projection with performance data of the tire; and training a machine learning model using the learning dataset so that, when data including image data of a target tire acquired based on the image acquisition conditions is input, performance data of the target tire is output. How to generate a trained model. [Explanation of symbols]

[0138] 21 Data 21A Tire image data 22 Performance Data 24B trained model

Claims

1. a derivation unit that inputs data including image data of a target tire acquired based on the image acquisition conditions into a trained model that has been machine-learned so as to output performance data of the tire when data including image data of the tire acquired based on the same predetermined perspective or projection image acquisition conditions is input, and outputs the performance data of the target tire. Tire performance prediction device.

2. The tire performance prediction device according to claim 1 , wherein the image data is acquired based on at least one of predetermined brightness conditions, size conditions, and image processing conditions in addition to the image acquisition conditions.

3. the data includes image data of a plurality of tires acquired based on the image acquisition conditions, The tire performance prediction device according to claim 1 , wherein the performance data includes a relative performance value based on the performance of one tire of the plurality of tires.

4. The tire performance prediction device according to claim 1 , wherein the image data includes a contour of a tread portion.

5. The tire performance prediction device according to claim 1 , wherein the image data includes a profile of a sidewall portion.

6. The tire performance prediction device according to claim 1 , wherein the performance data includes a performance value relating to at least one of drainage performance, mud performance, snow performance, and ice performance of the tire.

7. The tire performance prediction device according to claim 6 , wherein the performance value is obtained based on at least one of predetermined road surface conditions and weather conditions.

8. The tire performance prediction device according to claim 1 , wherein the performance data includes a rating regarding the appearance of the tire.

9. The tire performance prediction device according to claim 8 , wherein the rating is given by a predetermined evaluator.

10. The tire performance prediction device according to claim 1 , wherein the derivation unit further outputs contribution data that visualizes features that contributed to the output of the performance data.

11. 1. A method for predicting tire performance executed by one or more processors, comprising: a step of inputting data including image data of a target tire acquired based on the same predetermined perspective or projection image acquisition conditions into a trained model that has been machine-learned so as to output performance data of the tire when data including image data of the tire acquired based on the same predetermined perspective or projection image acquisition conditions is input; and outputting performance data of the target tire from the trained model. A method for predicting tire performance.

12. preparing a learning dataset that combines data including tire image data acquired under predetermined image acquisition conditions of the same perspective or projection with performance data of the tire; and training a machine learning model using the learning dataset so that, when data including image data of a target tire acquired based on the image acquisition conditions is input, performance data of the target tire is output. How to generate a trained model.

Images