Image recognition method, image recognition device, and image recognition program

Abstract

To perform segmentation without using machine learning or with a relatively small amount of teacher data if machine learning is used.SOLUTION: A feature amount extraction unit 11 generates a base feature map group consisting of a plurality of base feature maps from an input image, and generates a plurality of statistical amount maps by conducting a plurality of kinds of statistical amount computation on the base feature maps in the base feature map group. An inference device 13-i derives a segmentation inference result in response to an inference input based on the plurality of statistical amount maps. Then, in each of the plurality of kinds of statistical amount computation, a statistical amount is computed by a specific window size and by a specific computation expression. The window size and / or the computation expression vary according to the plurality of kinds of statistical amount computation.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an image recognition method, an image recognition device, and an image recognition program. [Background technology]

[0002] In recent years, inference machines (such as classifiers) obtained through machine learning have been put to practical use.

[0003] Generally, such reasoners require a large amount of training data to obtain inference results with sufficient accuracy, and if there is a relatively small amount of training data, good inference results may not be obtained due to bias in the training data.

[0004] To mitigate the effects of such bias in training data, ensemble learning is sometimes used. In ensemble learning, multiple highly independent reasoners are used, and a single final inference result is obtained from the inference results of the multiple reasoners by majority voting or other methods.

[0005] On the other hand, in the field of image recognition, an image processing device applies a spatial filter to an input image to be recognized, which extracts specific shapes (such as lines) of multiple sizes and multiple directions, and detects specific shapes of a certain size facing a certain direction contained in the input image (see, for example, Patent Document 1).

[0006] Furthermore, a certain inspection device (a) uses a machine learning model to derive a judgment result as to whether or not an input image contains an abnormality, and (b) calculates the degree of association between an image containing an abnormality and the input image, and the degree of association between an image not containing an abnormality and the input image, and evaluates the reliability of the above-mentioned judgment result based on the degree of association (see, for example, Patent Document 2). [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Japanese Patent Application Laid-Open No. 2017-13375 [Patent Document 2] Japanese Patent Application Publication No. 2019-20138 Summary of the Invention [Problem to be solved by the invention]

[0008] For collective learning of multiple inference devices (such as classifiers) for image recognition, it is conceivable to perform machine learning of each inference device based on features indicating specific shapes detected as described above. However, it is difficult to prepare features that are necessary for collective learning as training data for multiple inference devices for image recognition, and that are highly independent and capable of outputting inference results with sufficient accuracy.

[0009] Furthermore, in cases where a human being confirms an anomaly after it has been detected, it is necessary to identify the location of the anomaly in the input image. In such cases, the location of the anomaly in the input image can be identified by segmentation. However, since good segmentation requires learning the location (area) of the anomaly, it is even more difficult to prepare sufficient training data for segmentation.

[0010] The present invention has been made in consideration of the above-mentioned problems, and aims to provide an image recognition method, an image recognition device, and an image recognition program that can perform segmentation without using machine learning, or even if machine learning is used, with a relatively small amount of training data. [Means for solving the problem]

[0011] An image recognition method according to the present invention includes a feature extraction step of generating a group of base feature maps from an input image, the group of base feature maps being composed of a plurality of base feature maps, and performing a plurality of types of statistical calculations on the base feature maps to generate a plurality of statistical maps, and an inference step of deriving a segmentation inference result with an inference unit for an inference input based on the plurality of statistical maps. The plurality of types of statistical calculations are processes for calculating statistics using a specific calculation formula with a specific window size, and at least one of the window size and the calculation formula is different between the plurality of types of statistical calculations. The system further comprises the following configuration (A) or (B): (A) further comprising an integration step, in which the computer derives a plurality of inference results for a plurality of inference inputs based on the plurality of statistical maps using a plurality of inference units, respectively, and in which the computer integrates the plurality of inference results in a predetermined manner to derive a final inference result, each of the plurality of inference inputs having a statistical map that is part or all of the statistical maps of the plurality of statistical maps, and each inference input in the plurality of inference inputs has a statistical map that is partly or entirely different from the statistical maps of other inference inputs in the plurality of inference inputs. (B) The method further includes an integration step and an inference input generation step, wherein in the inference step, the computer derives multiple inference results using multiple inference devices for multiple inference inputs based on the multiple statistical maps, and in the integration step, the computer integrates the multiple inference results in a predetermined manner to derive a final inference result. In the inference input generation step, the computer generates the multiple inference inputs from the multiple statistical maps, and the multiple base feature maps are extracted from the input image by multiple specific processes, respectively, and the inference input has one or multiple statistical maps selected from the multiple statistical maps corresponding to the multiple specific processes.

[0012] The image recognition device according to the present invention includes a feature extraction unit that generates a group of base feature maps consisting of a plurality of base feature maps from an input image, and performs a plurality of types of statistical calculations on the base feature maps in the group of base feature maps to generate a plurality of statistical maps; Multiple For inference input TeSe of segmentation Multiple The inference results each Derive Multiple The plurality of types of statistical quantity calculations are processes for calculating statistics using specific calculation formulas with specific window sizes, and at least one of the window sizes and the calculation formulas differs among the plurality of types of statistical quantity calculations. The apparatus further includes the following configuration (A) or (B): (A) further including an integrator that integrates the multiple inference results in a predetermined manner to derive a final inference result, the multiple inference inputs each having a statistical map of some or all of the multiple statistical maps, and each inference input in the multiple inference inputs has a statistical map that is partially or completely different from the statistical maps of other inference inputs in the multiple inference inputs; (B) further including an integrator and an inference input generation unit, the multiple inference units derive multiple inference results for multiple inference inputs based on the multiple statistical maps, the integrator integrates the multiple inference results in a predetermined manner to derive a final inference result, the inference input generation unit generates the multiple inference inputs from the multiple statistical maps, the multiple base feature maps are extracted from the input image by multiple specific processes, respectively, and the inference input has one or more statistical maps selected from the multiple statistical maps corresponding to the multiple specific processes.

[0013] The image recognition program according to the present invention includes a computer including a feature extraction unit that generates a group of base feature maps from an input image, the group of base feature maps being made up of a plurality of base feature maps, and performs a plurality of types of statistical calculations on the base feature maps in the group of base feature maps to generate a plurality of statistical maps; Multiple For inference input TeSe of segmentation Multiple The inference results each Derive MultipleThe plurality of types of statistical quantity calculations are processes for calculating statistics using specific calculation formulas with specific window sizes, and at least one of the window sizes and the calculation formulas differs between the plurality of types of statistical quantity calculations. The present invention further includes the following configuration (A) or (B): (A) causing the computer to further function as an integrator, wherein the integrator integrates the multiple inference results in a predetermined manner to derive a final inference result, the multiple inference inputs each having a statistical map of some or all of the multiple statistical maps, and each inference input in the multiple inference inputs has a statistical map that is partially or completely different from the statistical maps of other inference inputs in the multiple inference inputs; (B) causing the computer to further function as an integrator and an inference input generation unit, wherein the integrator integrates the multiple inference results in a predetermined manner to derive a final inference result, the inference input generation unit generating the multiple inference inputs from the multiple statistical maps, the multiple base feature maps being extracted from the input image by multiple specific processes, respectively, and the inference input having one or more statistical maps selected from the multiple statistical maps corresponding to the multiple specific processes. [Effects of the Invention]

[0014] According to the present invention, an image recognition method, an image recognition device, and an image recognition program are provided that can perform segmentation without using machine learning, or even if machine learning is used, with a relatively small amount of training data.

[0015] The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings. [Brief explanation of the drawings]

[0016] [Figure 1] FIG. 1 is a block diagram showing the configuration of an image recognition device according to a first embodiment of the present invention. [Figure 2] FIG. 2 is a block diagram showing the configuration of the feature extraction unit 11 in FIG. [Figure 3] FIG. 3 is a diagram illustrating an example of the operation of the feature extraction unit 11 shown in FIG. [Figure 4] FIG. 4 is a diagram illustrating the operation of the statistics map derivation unit 23 in FIG. [Figure 5] FIG. 5 is a diagram illustrating an example of the operation of the inference input generating unit 12 in FIG. [Figure 6] FIG. 6 is a diagram illustrating clustering in the image recognition device according to the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0017] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0018] Embodiment 1

[0019] Fig. 1 is a block diagram showing the configuration of an image recognition device according to embodiment 1 of the present invention. The image recognition device shown in Fig. 1 is an electronic device such as a multifunction peripheral or scanner, a terminal device such as a personal computer, a server on a network, or the like, and executes an image recognition program on a built-in computer to cause the computer to function as a processing unit, which will be described later.

[0020] The image recognition device shown in FIG. 1 includes a feature extraction unit 11, an inference input generation unit 12, a plurality of inference units 13-1 to 13-N (N>1), an integrator 14, a weight setting unit 15, and a machine learning processing unit 16.

[0021] The feature extraction unit 11 generates a group of base feature maps consisting of a plurality of base feature maps from the input image, and performs a plurality of types of statistical calculations on the base feature maps in the group of base feature maps to generate a plurality of statistical maps.

[0022] The input image is an image that is the subject of image recognition, such as an image read by a scanner not shown, an image based on image data received by a communication device not shown, or an image based on image data stored in a storage device not shown.

[0023] The above-mentioned multiple base feature maps are extracted from the input image by multiple specific processes (here, spatial filtering processes). For example, several tens to several hundreds of base feature maps are generated and used as a single base feature map group.

[0024] Furthermore, the statistics map indicates the values of the results of statistical calculations (average, variance, etc.) at each pixel position.

[0025] Fig. 2 is a block diagram showing the configuration of the feature extraction unit 11 in Fig. 1. Fig. 3 is a diagram illustrating an example of the operation of the feature extraction unit 11 shown in Fig. 2.

[0026] 2, the feature extraction unit 11 includes a filter unit 21, a filter output integrating unit 22, and a statistics map derivation unit 23. The filter unit 21 performs filtering on the input image using a plurality of spatial filters with predetermined characteristics, and the filter output integrating unit 22 generates a base feature map based on the results of the filtering performed by the filter unit 21 at each position of the input image.

[0027] For example, as shown in Figure 3, to detect specific shapes (lines such as straight lines and curves, points, circles, polygons, etc.), multiple spatial filters with different detection sensitivities depending on the direction are used for each of multiple sizes, and a base feature map containing the shape of the logical sum of the filter outputs of the multiple spatial filters is generated. For example, if a line shape appears only in the filter output of a certain spatial filter and no shape appears in the filter outputs of any other spatial filters, a base feature map containing that line shape is generated. Also, if a line shape appears in the filter outputs of multiple spatial filters, a base feature map containing the points where the line shapes intersect (i.e., the point shape that is the logical product of the line shapes) is generated.

[0028] For example, a two-dimensional Gabor filter is used as this spatial filter. In this case, a two-dimensional Gabor filter with filter characteristics that match the spatial frequency corresponding to the size of the detection target is used. Alternatively, a second-order differential spatial filter that detects the edge of a shape may be used as this spatial filter.

[0029] Here, the base feature map has two-dimensional data indicating the positions, sizes, and orientations of multiple specific shapes. For example, the multiple specific shapes are detected in the input image by the spatial filtering process described above as the specific process. The base feature map may also be image data of specific colors (each color plane) in the input image. In this way, the base feature map with shape information and the base feature map with color information are used as needed.

[0030] Fig. 4 is a diagram illustrating the operation of the statistics map derivation unit 23 in Fig. 3. For example, as shown in Fig. 4, the statistics map derivation unit 23 performs a plurality of types of statistics calculations on the base feature map, thereby generating a statistics map.

[0031] The above-mentioned multiple types of statistical calculations are each a process of calculating statistics using a specific window size (the vertical and horizontal pixel size of a window centered on a pixel of interest) and a specific calculation formula (a calculation formula for a specified statistical quantity such as the average or variance), and at least one of the window size and the calculation formula differs between the multiple types of statistical calculations.

[0032] For example, if an input image containing text contains a linear abnormal object, the number, size, density, etc. of local lines will differ between the text and the abnormal object, so by utilizing various spatial statistics, it will be possible to detect abnormal areas (i.e., cases where an abnormal object is present).

[0033] That is, the statistics map derivation unit 23 performs a filter operation process for each type of statistics calculation, deriving statistics such as the mean and variance using a specified window size, on the base feature map, one pixel at a time, to generate a statistics map of the same size (same number of pixels vertically and horizontally) as the base feature map. Note that the statistics map derivation unit 23 may perform the above-mentioned filter operation process at intervals of N pixels (N>1) (i.e., perform the process on only one pixel out of N pixels) to generate a statistics map. In this case, the number of pixels in the statistics map is reduced, thereby reducing the amount of calculation in subsequent processing.

[0034] 1, the inference input generation unit 12 generates an inference input from the group of statistics maps (the plurality of statistics maps described above). In this embodiment, the plurality of inference inputs are input data input to the inference units 13-1 to 13-N, respectively.

[0035] Each of the plurality of inference inputs has a statistical map that is some or all of the plurality of statistical maps described above, and each inference input in the plurality of inference inputs has a statistical map that is some or all different from the statistical maps of other inference inputs in the plurality of inference inputs.

[0036] Furthermore, the above-mentioned multiple base feature maps are extracted from the input image by multiple specific processes, respectively, and each inference input has one or multiple statistical maps selected from the multiple statistical maps corresponding to the multiple specific processes.

[0037] Note that one of the above-mentioned multiple inference inputs may include all of the base feature maps in the group of base feature maps.

[0038] For example, each of the multiple inference inputs described above has one or more statistical maps selected corresponding to the multiple specific processes described above. That is, a certain inference input is composed of only the statistical maps generated from the base feature map obtained by a certain specific process among all the statistical maps.

[0039] Here, the base feature map has two-dimensional data indicating the position, size, and orientation of multiple specific shapes, and the multiple inference inputs are one or more statistical maps sorted by their sizes.

[0040] FIG. 5 is a diagram illustrating an example of the operation of the inference input generation unit 12 in FIG. 1. For example, as shown in FIG. 5, the above-mentioned multiple inference inputs are, for example, one or more statistical maps classified by size. Specifically, multiple size ranges are set, and for each size range, one or more statistical maps (hereinafter referred to as a statistical map set) in which the size of a specific shape belongs to that size range are treated as one inference input. In other words, here, classification is performed by size, not by position or method. Note that each size range may overlap partly or entirely with other size ranges.

[0041] Each inference input may also include data other than one or more statistical maps selected from the statistical maps (metadata such as parameters that may affect the inference result.) Examples of such metadata include environmental data at the time of image capture (temperature, humidity, time, status information of the subject, etc.; for example, if the input image is a photograph taken with a camera, environmental data at the time of capture), knowledge information (position and size of the area of interest), etc.

[0042] The above-mentioned plurality of inference inputs may be one or more statistical maps classified by one or both of the window size and the calculation formula in the statistical calculation.

[0043] Returning to Figure 1, the inference unit 13-i (i = 1,...,N) derives segmentation inference results (such as classification results of whether or not there is an abnormality at each pixel position) for multiple inference inputs based on the multiple statistical maps described above.

[0044] Specifically, in embodiment 1, multiple inference units 13-i each derive multiple inference results for multiple inference inputs based on the above-mentioned multiple statistical maps, and the integrator 14 integrates the multiple inference results in a predetermined manner to derive a final inference result.

[0045] In the first embodiment, the inference unit 13-i is an inference unit that has undergone machine learning. Note that the training data used for the machine learning of the inference unit 13-i is an input image that can obtain a base feature map in which the positions and directions of the specific shapes described above are distributed in all directions without bias.

[0046] In the first embodiment, the inference unit 13-i is a processing unit that derives an inference result for an inference input based on the above-described base feature map group, and is a processing unit that has undergone machine learning such as deep learning. For example, each of the inference units 13-i (i=1, . . . , N) is a convolutional neural network (CNN). For example, the multiple inference units 13-1 to 13-N are three or more inference units.

[0047] The integrator 14 is a processing unit that integrates the multiple inference results obtained by the multiple inference devices 13-1 to 13-N using a predetermined method (majority vote, class belonging probability, etc.) to derive a final inference result. For example, the integrator 14 derives the final inference result by majority vote on the multiple inference results, or based on the average or total value of class belonging probability for multiple classes (e.g., presence or absence of anomaly) for the multiple inference results.

[0048] In this embodiment, the integrator 14 integrates the multiple inference results in a predetermined manner, taking into account weighting factors for the multiple inference results, to derive a final inference result. Note that the final inference result may also be derived by integration without taking the weighting factors into account. The more reliable the inference result, the larger the weighting factor.

[0049] The integrator 14 may be an integrator that has undergone machine learning and may integrate the above-described multiple inference results to derive a final inference result. Alternatively, the integrator 14 may integrate the above-described multiple inference results using another existing method to derive a final inference result.

[0050] The weight setter 15 is a processing unit that derives and sets the above-mentioned weight coefficients in the integrator 14. The values of the weight coefficients may be set based on manually input values, or may be set automatically as follows.

[0051] For example, the weight setter 15 may derive the above-mentioned weight coefficients based on the inference accuracy of each of the multiple inference devices 13-1 to 13-N and set them in the integrator 14. In this case, for example, the machine learning processing unit 16 described below may derive the inference accuracy of each inference device 13-i by cross-validation (a verification method in which a process of dividing teaching data, using a portion for machine learning to derive an inference result, and using the remainder for verifying the inference result is repeated while changing the division pattern), and the weight setter 15 may derive the above-mentioned weight coefficients for the inference results of the multiple inference devices 13-1 to 13-N based on the inference accuracy of the multiple inference devices 13-1 to 13-N derived by the machine learning processing unit 16.

[0052] In this case, the inference accuracy of each inference unit 13-i may be estimated from the input image using an image recognition algorithm using, for example, CNN, etc. Furthermore, for example, the weight setter 15 may derive the weight coefficients described above based on the distribution of specific features (shape, color, etc.) for the input image and the distribution of specific features for the input images of the training data used in the machine learning of the multiple inference units 13-1 to 13-N, and set the weight coefficients in the integrator 14.

[0053] The machine learning processing unit 16 is a processing unit that executes a machine learning step of performing machine learning on the multiple inference units 13-1 to 13-N in accordance with an existing learning method corresponding to the computation model (here, CNN) of the inference units 13-1 to 13-N. In the machine learning on the multiple inference units 13-1 to 13-N, the machine learning on each inference unit 13-i is executed independently.

[0054] Specifically, training data including multiple pairs of input images and final inference results is prepared in a storage device (not shown), and the machine learning processing unit 16 acquires the training data, inputs each pair of input images to the feature extraction unit 11, acquires the inference results output from the inference units 13-1 to 13-N corresponding to the input images, and adjusts the parameter values (CNN weights and bias values) of each inference unit 13-i independently of the other inference units 13-j based on the results of comparing the output inference results with the final inference results for the training data pair.

[0055] The machine learning processing unit 16 may perform machine learning by excluding areas other than the specific partial area specified by the training data in the input image of the training data used for the above-mentioned machine learning. In other words, in this case, areas that require attention in image recognition (such as areas where specific parts of a machine or an area where an abnormality to be detected by image recognition may occur) are specified as specific partial areas, and other areas are excluded for machine learning, thereby efficiently progressing machine learning. For example, by limiting the area to areas where a specific abnormality to be detected by image recognition may occur and extracting a base feature map of a specific shape corresponding to the abnormality, machine learning can be performed efficiently with a relatively small amount of training data.

[0056] It should be noted that if the machine learning of the inference units 13-1 to 13-N has been completed, the machine learning processing unit 16 does not need to be provided.

[0057] Next, the operation of the image recognition device according to the first embodiment will be described.

[0058] (a) Machine learning of inference units 13-1 to 13-N

[0059] As training data, a plurality of pairs of input images and final inference results (i.e., correct image recognition results) are prepared in a storage device (not shown), etc. Then, machine learning processing unit 16 uses the training data to perform machine learning on inference devices 13-1 to 13-N.

[0060] In machine learning, when the machine learning processing unit 16 selects one piece of training data and inputs one input image of the training data to the feature extraction unit 11, the feature extraction unit 11 generates a group of statistics maps from the input image, and the inference input generation unit 12 generates each inference input from the group of statistics maps and inputs it to each inference device 13-i. Then, the inference devices 13-1 to 13-N each derive an inference result for the inference input based on their current state (such as CNN parameter values). Then, the machine learning processing unit 16 compares the inference result corresponding to the input image of the training data with the final inference result of the training data and updates the state of each inference device 13-1 to 13-N based on the comparison result using a predetermined algorithm.

[0061] In machine learning, this series of processes is repeatedly executed according to a predetermined machine learning algorithm depending on the values of hyperparameters such as the number of epochs.

[0062] (b) Image recognition (segmentation) of the input image to be recognized

[0063] After the above-described machine learning, image recognition is performed on the input image to be recognized. At this time, the input image (input image data) acquired by a controller (not shown) or the like is input to the feature extraction unit 11. When the input image is input to the feature extraction unit 11, the feature extraction unit 11 generates a group of statistical maps from the input image, and the inference input generation unit 12 generates each inference input from the group of statistical maps and inputs them to each inference unit 13-i. Then, the inference units 13-1 to 13-N each derive an inference result for the inference input based on the machine learning state (such as CNN parameter values). The integrator 14 then derives and outputs a final inference result from these inference results. The final inference result is a two-dimensional map indicating the degree of abnormality at each pixel position.

[0064] As described above, according to the first embodiment, the feature extraction unit 11 generates a group of base feature maps from an input image, which includes a plurality of base feature maps, and generates a plurality of statistical maps by performing a plurality of statistical calculations on the base feature maps in the group of base feature maps. The inference unit 13-i derives a segmentation inference result for an inference input based on the plurality of statistical maps. The plurality of statistical calculations is a process of calculating statistics using a specific window size and a specific calculation formula, and at least one of the window size and the calculation formula differs among the plurality of statistical calculations.

[0065] As a result, multiple base feature maps showing various feature quantities are generated from the input image, and further, a combination of statistical quantity maps showing various statistical quantities of the multiple base feature maps is used as an inference input, and an inference result for segmentation is obtained in the inference unit 13-i. Therefore, even when machine learning is used, good segmentation can be performed with a relatively small amount of training data.

[0066] In addition, since good inference results can be obtained with a relatively small amount of training data, good inference results suitable for the site can be obtained even in cases where there is little training data at individual, small-scale sites requiring image recognition. Furthermore, the statistical map visualizes the input of each inference unit 13-i, making it easy to explain the input-output relationship of each inference unit 13-i.

[0067] In this way, a base feature map showing features such as color, direction, and spatial frequency (object size) is generated in response to processing in the V1 area of the human visual cortex, and a statistical map is generated in response to subsequent higher-level processing in the human visual cortex, making it possible to perform general-purpose image recognition (in this case, anomaly detection) using a method similar to human image recognition.

[0068] Embodiment 2

[0069] In the second embodiment, an inference unit that generates an inference result by clustering without using machine learning is used instead of the inference units 13-1 to 13-N, the integrator 14, the weight setter 15, and the machine learning processing unit 16. In other words, in the second embodiment, machine learning is not required.

[0070] FIG. 6 is a diagram illustrating clustering in an image recognition device according to a second embodiment. For example, in the second embodiment, (a) from all statistical maps, statistical maps corresponding to the same identification process (the spatial filter process described above), the same window size, and the same statistical calculation formula are extracted as inference inputs by the inference input generation unit 12. (b) For each pixel position or subregion position of a predetermined size, feature values (e.g., mean and variance) indicated by the statistical maps are plotted in a feature space (a two-dimensional space of mean and variance in FIG. 2 ). (c) Among these plots, plots with a Mahalanobis distance greater than a predetermined value are determined to be abnormal areas, and the positions of the plots are identified as the positions of the abnormal areas. In this way, segmentation of the abnormal areas is performed. Note that, although the feature space in FIG. 2 is a two-dimensional space based on two feature values, it may also be a three-dimensional or higher-dimensional space based on three or more feature values.

[0071] The other configurations and operations of the image recognition device according to the second embodiment are the same as those of the first embodiment, and therefore the description thereof will be omitted.

[0072] As described above, according to the second embodiment, good segmentation can be performed without using machine learning.

[0073] It should be noted that various changes and modifications to the above-described embodiments will be apparent to those skilled in the art. Such changes and modifications may be made without departing from the spirit and scope of the subject matter and without diminishing its intended advantages. In other words, it is intended that such changes and modifications be included within the scope of the claims.

[0074] For example, in the above-mentioned embodiment 1, each of the inference units 13-1 to 13-N may be provided with a multi-layer inference unit, and each inference unit 13-i may derive an inference result using the multi-layer inference unit according to the stacking method of ensemble learning.

[0075] Furthermore, in the above-mentioned embodiment 1, when the above-mentioned metadata is input to the inference units 13-1 to 13-N, the same metadata may be input to the inference units 13-1 to 13-N, or (different from each other) metadata corresponding to each inference unit 13-i may be input to the inference units 13-1 to 13-N. [Industrial Applicability]

[0076] The present invention is applicable to, for example, image recognition. [Explanation of symbols]

[0077] 11 Feature extraction unit 12 Inference Input Generation Unit 13-1~13-N Reasoner

Claims

1. A feature extraction step of generating, by a computer, a group of base feature maps consisting of a plurality of base feature maps from an input image, and performing, by the computer, a plurality of types of statistical calculations on the base feature maps in the group of base feature maps to generate a plurality of statistical maps; an inference step of deriving an inference result of segmentation by an inference unit for an inference input based on the plurality of statistical maps; an integration step; the plurality of types of statistical calculations are processes for calculating statistical quantities using specific calculation formulas with specific window sizes, at least one of the window size and the calculation formula is different among the plurality of types of statistical quantity calculations; In the inference step, the computer derives a plurality of inference results using a plurality of inference units for a plurality of inference inputs based on the plurality of statistical quantity maps, respectively; In the integration step, the computer integrates the plurality of inference results in a predetermined manner to derive a final inference result; each of the plurality of inference inputs has a statistical map of some or all of the plurality of statistical maps; each inference input in the plurality of inference inputs has a statistical map that is partly or entirely different from the statistical maps of other inference inputs in the plurality of inference inputs; An image recognition method characterized by:

2. A feature extraction step of generating, by a computer, a group of base feature maps consisting of a plurality of base feature maps from an input image, and performing, by the computer, a plurality of types of statistical calculations on the base feature maps in the group of base feature maps to generate a plurality of statistical maps; an inference step of deriving an inference result of segmentation by an inference unit for an inference input based on the plurality of statistical maps; an integration step; an inference input generating step; the plurality of types of statistical calculations are processes for calculating statistical quantities using specific calculation formulas with specific window sizes, at least one of the window size and the calculation formula is different among the plurality of types of statistical quantity calculations; In the inference step, the computer derives a plurality of inference results using a plurality of inference units for a plurality of inference inputs based on the plurality of statistical quantity maps, respectively; In the integration step, the computer integrates the plurality of inference results in a predetermined manner to derive a final inference result; In the inference input generating step, the computer generates the plurality of inference inputs from the plurality of statistical quantity maps; the plurality of base feature maps are extracted from the input image by a plurality of specific processes, respectively; the inference input has one or more statistical maps selected from the plurality of statistical maps corresponding to the plurality of specific processes; An image recognition method characterized by:

3. the base feature map has two-dimensional data indicating positions, sizes, and orientations of a plurality of specific shapes; the inference input being one or more statistical maps sorted by size; 3. The image recognition method according to claim 1 or 2, wherein:

4. 3. The image recognition method according to claim 1, wherein the inference device is an inference device that has undergone machine learning.

5. 3. The image recognition method according to claim 1, wherein the inference unit generates an inference result by clustering without using machine learning.

6. a feature extraction unit that generates a group of base feature maps from an input image, the group of base feature maps being composed of a plurality of base feature maps, and performs a plurality of types of statistical calculations on the base feature maps in the group of base feature maps to generate a plurality of statistical maps; a plurality of inference units that derive a plurality of segmentation inference results for a plurality of inference inputs based on the plurality of statistical maps; an integrator; the plurality of types of statistical calculations are processes for calculating statistical quantities using specific calculation formulas with specific window sizes, at least one of the window size and the calculation formula is different among the plurality of types of statistical quantity calculations; the integrator integrates the plurality of inference results in a predetermined manner to derive a final inference result; each of the plurality of inference inputs has a statistical map of some or all of the plurality of statistical maps; each inference input in the plurality of inference inputs has a statistical map that is partly or entirely different from the statistical maps of other inference inputs in the plurality of inference inputs; An image recognition device characterized by the above.

7. A feature extraction unit that generates a base feature map group consisting of a plurality of base feature maps from an input image, and performs a plurality of types of statistical calculations on the base feature maps in the base feature map group to generate a plurality of statistical maps; a plurality of inference units that derive a plurality of segmentation inference results for a plurality of inference inputs based on the plurality of statistical maps; an integrator; an inference input generation unit; the plurality of types of statistical calculations are processes for calculating statistical quantities using specific calculation formulas with specific window sizes, at least one of the window size and the calculation formula is different among the plurality of types of statistical quantity calculations; the integrator integrates the plurality of inference results in a predetermined manner to derive a final inference result; the inference input generation unit generates the plurality of inference inputs from the plurality of statistical quantity maps; the plurality of base feature maps are extracted from the input image by a plurality of specific processes, respectively; the inference input has one or more statistical maps selected from the plurality of statistical maps corresponding to the plurality of specific processes; An image recognition device characterized by the above.

8. Computer, a feature extraction unit that generates a group of base feature maps from an input image, and performs a plurality of types of statistical calculations on the base feature maps in the group of base feature maps to generate a plurality of statistical maps; a plurality of reasoners that derive a plurality of inference results of segmentation for a plurality of inference inputs based on the plurality of statistical maps, respectively; and It acts as an integrator, the plurality of types of statistical calculations are processes for calculating statistical quantities using specific calculation formulas with specific window sizes, at least one of the window size and the calculation formula is different among the plurality of types of statistical quantity calculations; the integrator integrates the plurality of inference results in a predetermined manner to derive a final inference result; each of the plurality of inference inputs has a statistical map of some or all of the plurality of statistical maps; each inference input in the plurality of inference inputs has a statistical map that is partly or entirely different from the statistical maps of other inference inputs in the plurality of inference inputs; An image recognition program that features:

9. A computer, a feature extraction unit that generates a group of base feature maps from an input image, and performs a plurality of types of statistical calculations on the base feature maps in the group of base feature maps to generate a plurality of statistical maps; a plurality of inference units that respectively derive a plurality of segmentation inference results for a plurality of inference inputs based on the plurality of statistical maps; integrators, and Functioning as an inference input generator, the plurality of types of statistical calculations are processes for calculating statistical quantities using specific calculation formulas with specific window sizes, at least one of the window size and the calculation formula is different among the plurality of types of statistical quantity calculations; the integrator integrates the plurality of inference results in a predetermined manner to derive a final inference result; the inference input generation unit generates the plurality of inference inputs from the plurality of statistical quantity maps; the plurality of base feature maps are extracted from the input image by a plurality of specific processes, respectively; the inference input has one or more statistical maps selected from the plurality of statistical maps corresponding to the plurality of specific processes; An image recognition program that features:

Images