JP2016053751A - Information processing device, information processing method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily prepare test data for each software component.SOLUTION: A processing model difference detection unit 101 detects a difference in processing models between a reference processing model in which test data associated to a processing model has been prepared and a derivation processing model for which test data associated to the processing model has not been prepared. A retrieval unit 102 retrieves a test data conversion processing method from dictionary means by using the detected difference in the processing models as a retrieval key. A conversion processing unit 103 converts the test data of the reference processing model into test data associated to the derivation processing model on the basis of the retrieved conversion processing method.SELECTED DRAWING: Figure 1

Description

  The present invention particularly relates to an information processing apparatus, an information processing method, and a program suitable for use in creating test data corresponding to a newly created processing model from test data of a processing model that has already been managed in association. .

  Conventionally, as a test data utilization method, for example, a technique described in Patent Document 1 is known. Specifically, Patent Document 1 discloses a system test specification generation device that generates a system test specification that is executed in a state in which software components are combined with software configured by a plurality of software components. ing.

JP 2011-118637 A

  However, in the technique described in Patent Document 1, although the test of the entire system can be automatically generated, the test specification for each software component needs to be individually created by the user. For this reason, when creating a test specification for each component, there is a problem that the user's labor is large when creating test data.

  An object of the present invention is to make it possible to easily create test data for each software component in view of the above-described problems.

  The information processing apparatus according to the present invention corresponds to the newly created second processing model using the first processing model in which test data is managed in association with the processing model configured by the processing modules. An information processing apparatus for creating test data, the detecting means for detecting a difference between the first processing model and the second processing model, a difference between the first processing model and a test data conversion process A search unit that searches for a dictionary corresponding to the difference detected by the detection unit from a dictionary that holds the method in association with each other, and a test related to the first processing model using the dictionary searched by the search unit Conversion means for converting data into test data according to the second processing model.

  According to the present invention, test data can be easily created for each software component.

It is a block diagram which shows the function structural example of the test data conversion process performed in the test data conversion tool in 1st Embodiment. It is a figure which shows an example of the system configuration | structure which concerns on 1st Embodiment, and an example of figure programming. It is a figure which shows an example of the reference | standard process model in 1st and 4th embodiment. It is a figure for demonstrating the outline | summary of the face detection process in the reference | standard process model of 1st Embodiment. It is a figure which shows an example of the derived process model in 1st Embodiment. It is a figure for demonstrating the outline | summary of the face detection process in the derivation process model of 1st Embodiment. It is a figure which shows an example of the list | wrist of the processing module in 1st Embodiment, and a joint list | wrist. It is a figure which shows the specific structural example of the dictionary means in 1st Embodiment. It is a figure which shows an example of the conversion process model in 1st Embodiment. It is a figure for demonstrating the outline | summary of the filter process in the reference | standard process model of 2nd Embodiment. It is a figure which shows an example of the list | wrist of the processing module in 2nd Embodiment. It is a figure which shows the specific structural example of the dictionary means in 2nd Embodiment. It is a figure which shows an example of the conversion process model in 2nd Embodiment. It is a figure which shows an example of the list | wrist of the processing module in 3rd Embodiment. It is a figure which shows the specific structural example of the dictionary means in 3rd Embodiment. It is a figure which shows an example of the conversion process model in 3rd Embodiment. It is a figure for demonstrating the outline | summary of the face detection process in the derivative process model of 4th Embodiment. It is a figure which shows an example of the list of the processing modules in 4th Embodiment. It is a figure which shows the specific structural example of the dictionary means in 4th Embodiment. It is a figure which shows an example of the conversion process model in 4th Embodiment. It is a figure for demonstrating the outline | summary of the face detection process in the derivative process model of 5th Embodiment. It is a figure which shows an example of the list of the processing modules in 5th Embodiment. It is a figure which shows the specific structural example of the dictionary means in 5th Embodiment. It is a figure which shows an example of the conversion process model in 5th Embodiment. It is a flowchart which shows an example of the process sequence in the test conversion processing tool in 6th Embodiment. It is a figure which shows an example of the display screen for designating the conversion process model of test data newly and registering in a dictionary means.

(First embodiment)
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. The present embodiment assumes a processing model in which a plurality of processing modules having input / output terminals are connected in a predetermined arrangement or topology. Hereinafter, a test data conversion tool for creating test data corresponding to a newly created processing model from test data of the processing model already managed in association with the processing model will be described. Hereinafter, a process model for which association management is performed for newly creating a similar process model as described later from a process model for which test data is associated and managed is referred to as a reference process model. A newly created processing model is called a derived processing model.

First, in order to describe the test data conversion tool more specifically, a system configuration for executing this tool will be described.
FIG. 2A is a diagram illustrating a system configuration example according to the present embodiment.
As shown in FIG. 2 (a), the system for executing the test data conversion tool comprises a personal computer (PC) 4 to which a monitor 5 is connected and a database server 7 to which a database 6 is connected via a network. Has been. A graphic programming application software is installed in the PC 4 as the information processing apparatus, and a graphic programming process is presented to the user from the monitor 5. The database 6 holds information on the reference processing model and test data, and the database server 7 manages the database 6.

  The test data conversion tool in this embodiment is provided as a part of this graphic programming application software. The test data conversion tool may be provided as an independent tool, not as a part of the graphic programming application software.

  Here, the graphic programming will be briefly described. As shown in FIG. 2B, in the graphic programming in the present embodiment, the processing module 8 is connected via the input terminal 9 and the output terminal 10 on the monitor 5 by the user operating the mouse or the like. A processing model that realizes a desired function is created. The processing model is set with an ID as an attribute for identifying each, and each processing module included in the processing model is set to execute a predetermined process on the input data. Is equivalent to a program expressed in code.

FIG. 1 is a block diagram illustrating a functional configuration example of test data conversion processing executed by the test data conversion tool of the PC 4 according to the present embodiment.
In FIG. 1, the process model difference detection unit 101 is a process model difference between a reference process model for which test data corresponding to the process model has been created and a derived process model for which test data corresponding to the process model has not been created. Is detected. The search unit 102 searches the dictionary means for a test data conversion processing method using the detected processing model difference as a search key. The conversion processing unit 103 converts the test data of the reference processing model into test data corresponding to the derived processing model based on the searched conversion processing method.

  Here, before explaining the details of the test data conversion process executed in the test data conversion tool, the process model will be described.

  FIG. 3 is a diagram illustrating an example of the reference processing model in the present embodiment. For example, the processing model shown in FIG. 3 includes two processing modules, a filter processing module 301 and a face detection processing module 302. Image data is input to the filter processing module 301, and after the filter processing is performed on the image data, the image data to which the filter processing is applied is output.

  The image data to which the filter processing is applied is then input to the face detection processing module 302, and processing result data to which the face detection processing is applied is output from the face detection processing module 302. That is, as a whole reference processing model shown in FIG. 3, a filtering process and a face detection process are executed on the input image data, and a process of outputting process result data to which both processes are applied is executed.

  Here, the processing executed in the face detection processing module 302 described above will be described in more detail. The face detection processing module 302 has a function of detecting a human face included in an image. In the case of the present embodiment, for example, as shown in FIG. 4A, face detection processing is applied to a small area 401 of 21 × 21 pixels in an image 402. Further, the raster detection is performed while shifting the face detection process for the small area 401 from the upper left to the lower right of the image by shifting one pixel at a time. In the following description, it is assumed that the size of the image 402 as input data is 200 × 150 pixels. It is assumed that the image size changes corresponding to the processing model and processing module.

  When raster scanning the small area 401 in the image, face detection processing is sequentially executed by a double loop using a for sentence in C language, for example. Further, regarding the processing content of the face detection processing itself, a known method is used, and detailed description thereof is omitted.

  In the face detection processing module 302, as described above, the face detection process is sequentially executed on the input image by raster scanning, and when a human face is detected, the center coordinate data in the small area image is converted into an array format. Store sequentially in output data. For example, as shown in FIG. 4B, when a face is detected at four locations in the image, the face detection processing module 302 performs raster scanning in the image, so the addresses at the four locations are The sequence data is stored in the order shown in FIG. It is assumed that a processing module attribute “(1) raster scan” is set in advance for the face detection processing module 302 by the creator.

  Next, test data that is managed in association with the reference processing model will be described. Generally, a test method using test data is adopted as a means for determining whether a certain program is implemented without error. In this test method, output data to be output is calculated in advance as correct answer data for certain input data. Then, whether or not there is an error in the program is determined based on whether or not the output data that is output by inputting the input data to the test target program matches the correct data.

  In the present embodiment, the test data is held in the database 6 together with the reference processing model, and an ID is set as an attribute to identify each test data. In addition, for the reference processing model in the present embodiment, the same data as the image data and output data shown in FIG. 4B is previously associated and managed as input data and correct data, respectively. And The correct answer data is not calculated by the processing model, but is calculated separately in advance.

  Specifically, the reference process model and the test data are associated and managed by setting the ID of the corresponding test data as attribute data for the reference process model held on the database 6. At this time, if image data is input to the reference processing model, the same result as the correct data is output as shown in FIG. 4B, so the reference processing model is implemented without error. Can be determined.

  In the present embodiment, the test data is described as a combination of a set of input data and correct data in order to simplify the description. On the other hand, in practice, it may be composed of a combination of a plurality of sets of input data and correct answer data according to the complexity of the reference processing model. As described above, it is assumed that the reference processing model described with reference to FIGS. 3 and 4 is managed in association with test data, and is further guaranteed to be implemented without error.

  Next, the derivation process model will be described. The derived process model in the present embodiment refers to a model newly created by the designer with reference to the reference process model to realize a similar process. Specifically, for example, the same face detection process is applied from the image data to the reference processing model described with reference to FIGS. 3 and 4 to detect a human face. The detection processing is executed in parallel.

  FIG. 5 is a diagram illustrating an example of the derivation processing model. As shown in FIG. 5, the derivation processing model in the present embodiment first includes the same filter processing module 301 as the reference processing model. Subsequently, in the derivation processing model, the filtered image data is input to the ten partial face detection processing modules 601 as shown in FIG.

  Here, as shown in FIG. 6A, each of the partial face detection processing modules 601a to 601j performs face detection with the pixel position of 10 pixels continuous in the vertical direction of the image data as the center position of the face detection subregion. Perform processing in parallel. In each of the partial face detection processing modules 601a to 601j, the result of detecting the face from the small area (coordinates on the image where the face is detected) is input to the data integration module 602.

  Note that it is assumed that the processing module attribute “(2) parallel in the column direction” is set in advance by the creator for the partial face detection processing module 601 in the derivation processing model shown in FIG. Further, it is assumed that the processing module attribute “(3) column direction integration” is set in advance by the creator for the data integration module 602 in the derived processing model. Furthermore, in this derivation processing model, it is assumed that the test data is not associated and managed.

  In the data integration module 602, the coordinate data of the pixel position where the face is detected is stored in the array data in the order of the partial face detection processing modules 601a to 601j (the order indicated by the arrows in FIG. 6A). . Note that the data integration module 602 has a function of sequentially storing input data in array data when the data input from the partial face detection processing module has a value other than NULL. If there is an input having a value other than NULL from a plurality of partial face detection processing modules, the partial face detection processing modules 601a to 601j are stored in the array data in the order.

  When the above processing is repeated for one vertical column of an image (in the case of the present embodiment, 150 pixels / 10 pixels = 15 times), the processing moves to the right adjacent column and the same processing is repeated. When the above processing is executed for all the image data columns (that is, when the face detection processing for the entire image is completed), the data integration module 602 outputs array data in which the detection results are stored.

  Here, for example, in the example shown in FIG. 6B, faces are detected at four locations in the image, but the order in which the face detection processing results are stored in the output data array in the derivation processing model is shown in FIG. The order is as shown in FIG. This order is different from the order shown in FIG. 4B, and it can be seen that the order is different from the raster scan direction in the reference processing model.

  Next, the processing procedure of each unit related to the test data conversion processing in the present embodiment shown in FIG. 1 will be described. As described above, the test data conversion tool in this embodiment is executed as a part of graphic programming application software. It is assumed that the user has created the derivation process model described above with respect to the reference process model that has been created in advance. In addition, it is assumed that the corresponding reference processing model is acquired in advance from the database 6 before performing the following processing.

  First, a test is executed to determine whether or not the derivation processing model is implemented without error. Here, since test data is not managed in association with the derivation process model, the test data conversion tool executes conversion processing on the test data of the reference process model to create test data corresponding to the derivation process model.

  In the test data conversion tool, first, the process model difference detection unit 101 detects a process model difference between the reference process model specified by the user and the derived process model. In the case of this embodiment, the reference processing model and the derived processing model are traced in order from the input terminal to the processing module, and a combined list of processing modules to be connected is created. FIG. 7C shows a combined list created from the reference processing model, and FIG. 7D shows a combined list created from the derived processing model.

  Here, according to the combined list in FIG. 7C, it can be seen that the filter processing module and the face detection processing module are connected in series. As processing model information, the processing module shown in FIG. Is created.

  On the other hand, according to the combined list of FIG. 7 (d), a plurality of partial face detection modules are connected after the filter processing module, and outputs of the plurality of partial face detection processing modules are connected to the data integration module. I understand that. Each partial face detection processing module is the same type of processing module, shares the output of one filter processing module as an input, and is coupled to one partial face detection processing module. From this, it is grouped as a group that executes the same processing. As a result, a list of processing modules shown in FIG. 7B is created as processing model information for the derived processing model.

  As shown in FIGS. 3 and 5, the respective processing models have different topologies and processing module attributes, and the difference between the processing models is detected as a difference between the processing module lists. That is, in this step, the difference between the two processing models is detected as follows by comparing the lists shown in FIGS. 7A and 7B.

  First, regarding the filtering module, since the data in both lists is the same, no difference is detected. Subsequently, the reference processing model has a face detection processing module, the processing category is detection processing, the detection target is a face, and the detection direction is set to raster scan.

  On the other hand, the derivation processing model has ten partial face detection processing modules, the processing category is detection processing, and the detection target is a face, which is the same as the reference processing model. However, unlike the reference processing model, the detection direction (processing module attribute) is set to the column direction scan. Further, it can be seen that the derived processing model includes one data integration module that the reference processing model does not have.

  Therefore, the processing categories of the face detection processing module and the partial face detection processing module are the same as the detection processing, and further, a difference in processing model is detected as a difference in processing module attributes, the number of processing modules, and the number of connected modules.

  Subsequently, in the test data conversion tool, processing by the search unit 102 shown in FIG. 1 is executed. The retrieval unit 102 retrieves a test data conversion processing method from the dictionary unit using the detected processing model difference and the processing module processing category detected as described above as search keys.

  Here, the dictionary means will be described. As shown in FIG. 8, the dictionary means in this embodiment is configured as a dictionary that holds differences in processing models as search keys and holds test data conversion processing methods corresponding to the differences in the respective processing models as values. The

  Here, the search key enumerates the label “detection process” that is common to the reference process model and the derived process model, and the process module attribute, the number of process modules, and the number of connected modules that are different in each process model described above. It is described as data. Further, as described above, the derivation processing model has a data integration module that the reference processing model does not have. However, in this respect as well, as shown in FIG. The search key fields are described in order.

  Note that the label “detection process” related to the process category described above is not a difference regarding each process model. However, the difference in the processing direction of the face detection processing shown in the present embodiment can be applied to the entire detection processing for images, and is added as information in order to make the search key versatile.

  Note that the dictionary means is held in a storage medium such as a hard disk (not shown) in the PC 4, but may be held in the database 6. Similarly to a general database, the dictionary means has a function as a dictionary that outputs a value (ID of a predetermined processing model) corresponding to a search key (difference in processing model) when it is input. Accordingly, in the search process, information on the ID of a predetermined process model corresponding to the difference between the process models can be obtained by inputting the difference between the process models to the dictionary unit as a search key.

  In the present embodiment, the information on the ID of the predetermined processing model held in the dictionary means is specifically the ID of the processing model having the function of executing the processing content for the test data associated with the reference processing model. It is a representation. A processing model in which an ID is registered as a dictionary value is defined as a processing model for inputting test data, outputting test data after conversion, and converting test data (hereinafter referred to as a conversion processing model). ing.

  As described above, when a difference between processing models (for example, processing module attributes, number of processing modules, number of connected modules) is input to the dictionary means as a search key, ID-abc matching the search key as shown in FIG. The registered conversion processing model is searched.

  Subsequently, in the test data conversion tool, processing by the conversion processing unit 103 shown in FIG. 1 is executed. The conversion processing unit 103 converts the test data of the reference processing model into test data corresponding to the derived processing model by using the conversion processing model corresponding to the searched ID.

  The ID-abc conversion processing model searched as described above is registered as a processing model including the array order conversion processing module 1001 as shown in FIG. The array order conversion processing module 1001 inputs array data corresponding to the output data of the reference processing model, and converts the order of the elements of the input array data into the order of the elements of the array data corresponding to the output data of the derivation processing model. To do. That is, the array order conversion processing module 1001 is designed to output correct data in the test data of the derived processing model when correct data in the test data associated with the reference processing model is input.

  Specifically, in the reference processing model, since the detection process is executed in the raster scan direction, as shown in FIG. 4B, the address data on the image indicating the detection position prioritizes the y coordinate. They are stored in sequence data in order. On the other hand, in the derivation processing model, since the detection process is executed with priority in the column direction, the address data on the image indicating the detection position is arranged in the order in which the x coordinate is prioritized as shown in FIG. Stored in the data.

  Therefore, in the conversion processing model in the present embodiment, the input address data stored in the array data is sorted in the order in which the x coordinate is prioritized so as to correspond to the column direction detection processing. As a result, in this embodiment, the conversion processing unit 103 converts the correct data in the test data associated with the reference processing model into correct data corresponding to the derived processing model as shown in FIG.

  That is, as described above, there is a reference processing model associated with the test data, and a derivation process that matches a search key (difference in processing model) registered in the dictionary in advance based on the reference processing model. A user creates a new model. In this case, the user can generate test data corresponding to the derived processing model from the test data corresponding to the reference processing model.

  As described above, when there is a processing model (reference processing model) associated with the test data in advance by the test data conversion tool in the present embodiment, a new test data is created for the newly created processing model. Time and effort is reduced. That is, when a user newly creates a processing model (derivative processing model) having a difference between processing models registered in the dictionary based on the processing model, test data for the derived processing model is automatically generated. Is possible.

(Second Embodiment)
Hereinafter, a second embodiment of the present invention will be described with reference to the drawings. The test data conversion tool according to the present embodiment is different in that the data size of the input terminal and the output terminal and the processing module attribute are detected as a difference between the reference processing model and the derived processing model. Accordingly, the test data conversion process is also different. Hereinafter, in the present embodiment, only differences from the first embodiment will be described, and description of common parts will be omitted. Further, the system configuration and the functional configuration of the PC according to the present embodiment are the same as those in the first embodiment, and thus description thereof is omitted.

  The reference processing model in the present embodiment is composed of one filter processing module. When image data is input, the filter processing module performs filter processing on the image data and then outputs image data to which the filter processing has been applied. That is, as the entire reference processing model, a filtering process is performed on the input image data, and a process of outputting processing result data to which the filtering process is applied is performed.

Here, the processing executed in the above-described filter processing module will be described in more detail. In the filter processing module, a filter operation represented by a 3 × 3 pixel kernel 1201 shown in FIG. In FIG. 10, a part of the input image data is shown. At this time, for example, as illustrated in FIG. 10B, when the filter processing is applied when some of the pixel values in the image 1202 are values of a to i, the filter calculation value at the pixel position having the pixel value of e E is calculated by the following equation (1).
E = −a−2b−c + g + 2h + i (1)

  The filter processing module executes the filter processing on the entire surface of the input image data, and outputs the processing result as the image data subjected to the filter processing. In this embodiment, in the filtering process for the pixels on the outer frame of the image, as shown in FIG. 10C, the kernel 1201 is displayed in the vicinity of the upper left corner of the image data in the image 1202 (FIG. 10C). The calculation is performed with the pixel value of the portion protruding from 0) being zero. The image size of the output image data is made to match the size of the input image data. Further, it is assumed that the processing module attribute “(4) Image unit processing” is set in advance by the creator for the filter processing module.

  Next, test data that is managed in association with the reference processing model will be described. In the test data in this embodiment, together with the image data of the input data, image data that is a calculation result when the above-described filter calculation by the kernel is applied to the image data is prepared as correct data. Note that the correct answer data is not calculated by the processing model, but is calculated in advance. In the reference processing model according to the present embodiment, it is assumed that the test data is associated and managed as described above, and that it is guaranteed that the test processing is implemented without error.

  Next, the derivation process model will be described. The derivation processing model in the present embodiment is composed of an 8-pixel unit filter processing module, and is similar to the above-described reference processing model in that the filter processing represented by Expression (1) is performed on image data. is there. On the other hand, the size of the input data is different. That is, in the reference processing model, the size of the input data is the image size, whereas in the derived processing model, the size of the input data is a 3 × 10 pixel size to be filtered.

  The 3 × 10 pixel size is an input pixel size necessary for executing the filter processing for 8 pixels, and the pixel size of the output data of the derivation processing model is 8 pixels. As for the filter processing for pixels on the outer frame of the image, the calculation is performed with the pixel value of the portion where the kernel protrudes from the image as 0, as in the reference processing model. That is, with respect to the input 3 × 10 pixel data, 0 is substituted as the pixel value of the portion that protrudes from the image. Further, it is assumed that the processing module attribute “(5) 8-pixel unit processing” is set in advance by the creator for the 8-pixel unit filter processing module in the derived processing model. In the derivation processing model, it is assumed that test data is not associated and managed.

  Next, the test data conversion tool in this embodiment will be described. Similar to the first embodiment, the test data conversion tool in this embodiment is executed as part of graphic programming application software. It is assumed that the user has created the derivation process model described above with respect to the reference process model that has been created in advance.

  First, in the test data conversion tool, the process model difference detection unit 101 detects a process model difference between the reference process model and the derived process model specified by the user. Similarly, in this embodiment, a list of processing modules to be connected is created. FIG. 11A and FIG. 11B show an example of a list of processing modules created from the reference processing model and the derived processing model.

  Here, as shown in FIG. 11A and FIG. 11B, both processing models have different processing module attributes and input and output data sizes. Detected as a list difference. Specifically, from FIG. 11A and FIG. 11B, first, (4) image unit processing and (5) 8-pixel unit processing are detected as differences in processing module attributes. Then, the image size and the 3 × 10 pixel size are detected as the difference in the input data size, and the image size and the 8-pixel size are detected as the difference in the output data size.

  Subsequently, in the test data conversion tool, processing by the search unit 102 shown in FIG. 1 is executed. Similar to the first embodiment, the search unit 102 searches the dictionary means for a test data conversion processing method using the difference between the processing models as a search key. When a processing model difference (processing module attribute, input and output data size) is input as a search key, as a test data conversion processing method matching the search key as shown in FIG. 12, conversion processing registered as ID-def The model is searched.

  Subsequently, in the test data conversion tool, processing by the conversion processing unit 103 shown in FIG. 1 is executed. The conversion processing unit 103 converts the test data of the reference processing model into test data corresponding to the derived processing model based on the searched conversion processing model.

  The ID-def conversion processing model searched as described above is registered as a processing model including the image data conversion processing module 1601 as shown in FIG. The image data conversion processing module 1601 inputs image data corresponding to the input data or output data of the reference processing model, and outputs image data corresponding to the input data or output data of the derivation processing model. In other words, the image data conversion processing module 1601 is designed to output test data of a derived processing model when test data associated with a reference processing model is input.

  Specifically, a process of cutting out data of 3 × 10 pixel size from the image data in the input test data is executed, and correct data of 1 × 8 pixel size is extracted from the correct image data in the input test data Execute (cut out) processing. In the cutting process, as shown in FIG. 13, the image located at the upper left of the image is cut out. As a result, in the present embodiment, the conversion processing unit 103 converts the input data and correct data in the test data associated with the reference processing model into input data and correct data corresponding to the derived processing model as shown in FIG. To do.

  As described above, when there is a processing model (reference processing model) associated with the test data in advance by the test data conversion tool in the present embodiment, a new test data is created for the newly created processing model. Time and effort is reduced. That is, when a user newly creates a processing model (derivative processing model) having a difference between processing models registered in the dictionary based on the processing model, test data for the derived processing model is automatically generated. Is possible.

(Third embodiment)
The third embodiment of the present invention will be described below with reference to the drawings. The test data conversion tool according to this embodiment is different in that the processing module attribute is detected as the presence or absence of padding processing as a difference between the reference processing model and the derived processing model. Accordingly, the test data conversion process is also different. Hereinafter, in the present embodiment, only differences from the first embodiment will be described, and description of common parts will be omitted. Further, the system configuration and the functional configuration of the PC according to the present embodiment are the same as those in the first embodiment, and thus description thereof is omitted.

  The reference processing model in the present embodiment is composed of one filter processing module. When the image data is input, the filter processing module performs the filter processing on the image data and then outputs the image data to which the filter processing is applied. That is, as a whole reference processing model, a filtering process is performed on input image data, and a process of outputting processing result data to which the filtering process is applied is performed.

  The outline of the filter processing by the filter processing module is the same as that of the second embodiment as shown in FIG. 10, and the arithmetic processing executed in the filter processing module is the same as the formula ( 1). The filter processing module executes the filter processing on the entire surface of the input image data, and outputs the processing result as the image data subjected to the filter processing. It is assumed that the processing module attribute “(6) Padding 0” is set in advance for the filter processing module by the creator. Further, the test data that is managed in association with the reference processing model is the same as in the second embodiment, and the reference processing model in this embodiment is managed in association with the test data and is implemented without error. It is guaranteed that

  Next, the derivation process model will be described. The derivation processing model in the present embodiment is composed of a non-padding filter processing module, and is similar in that the filter processing represented by Expression (1) is performed on image data as compared with the above-described reference processing model. On the other hand, it is different from the reference processing model in that the filtering process is not performed on the pixels on the outer periphery of the image when the kernel protrudes from the image. As a result, the image size of the output image data is reduced by two pixels for both the vertical size and the horizontal size with respect to the size of the input image.

  It is assumed that the processing module attribute “(7) no padding” is set in advance by the creator for the filter processing module without padding constituting the derivation processing model. In the derivation processing model, it is assumed that test data is not associated and managed.

  Next, the test data conversion tool in this embodiment will be described. Similar to the first embodiment, the test data conversion tool in this embodiment is executed as part of graphic programming application software. It is assumed that the user has created the derivation process model described above with respect to the reference process model that has been created in advance.

  First, in the test data conversion tool, the process model difference detection unit 101 detects a process model difference between the reference process model and the derived process model specified by the user. Similarly, in this embodiment, a list of processing modules to be connected is created. FIG. 14A and FIG. 14B show an example of a list of processing modules created from the reference processing model and the derived processing model.

  Here, as shown in FIGS. 14A and 14B, both processing models have different processing module attributes, and a difference between the processing models is detected as a difference between the processing module lists. In the case of the present embodiment, as shown in FIGS. 14A and 14B, a difference is detected as a processing module attribute.

  Subsequently, in the test data conversion tool, processing by the search unit 102 shown in FIG. 1 is executed. Similar to the first embodiment, the search unit 102 searches the dictionary means for a test data conversion processing method using the difference between the processing models as a search key. When the processing model difference (processing module attribute) is input as a search key, as shown in FIG. 15, a conversion processing model registered as ID-ghi is searched as a test data conversion processing method that matches the search key. .

  Subsequently, in the test data conversion tool, processing by the conversion processing unit 103 shown in FIG. 1 is executed. The conversion processing unit 103 converts the test data of the reference processing model into test data corresponding to the derived processing model based on the searched conversion processing model.

  The ID-ghi conversion processing model searched as described above is registered as a processing model including the image data conversion processing module 2101 as shown in FIG. The image data conversion processing module 2101 receives image data corresponding to the output data of the reference processing model, and outputs image data obtained by removing pixels on the outer periphery of the image from the input image data. That is, the image data conversion processing module 2101 is designed to output correct data in the test data of the derived processing model when correct data in the test data associated with the reference processing model is input.

  Specifically, a process of extracting (cutting out) image data obtained by removing a range of pixels on the outer periphery of the image from the input correct answer data is executed. As a result, in the present embodiment, the conversion processing unit 103 converts the correct answer data in the test data associated with the reference process model into correct answer data corresponding to the derived process model as shown in FIG.

  As described above, when there is a processing model (reference processing model) associated with the test data in advance by the test data conversion tool in the present embodiment, a new test data is created for the newly created processing model. Time and effort is reduced. That is, when a user newly creates a processing model (derivative processing model) having a difference between processing models registered in the dictionary based on the processing model, test data for the derived processing model is automatically generated. Is possible.

(Fourth embodiment)
The fourth embodiment of the present invention will be described below with reference to the drawings. The test data conversion tool according to this embodiment is different in that a processing module attribute is detected as a difference between the reference processing model and the derived processing model. Accordingly, the test data conversion process is also different. Hereinafter, in the present embodiment, only differences from the first embodiment will be described, and description of common parts will be omitted. Further, the system configuration and the functional configuration of the PC according to the present embodiment are the same as those in the first embodiment, and thus description thereof is omitted.

  The reference processing model in the present embodiment is the same processing model as the reference processing model shown in FIG. 3 in the first embodiment, and includes two processing modules, a filter processing module 301 and a face detection processing module 302. Has been. The test data associated and managed with respect to the reference processing model is also the same as in the first embodiment.

  Next, the derivation process model will be described. The derivation processing model in this embodiment includes a column-direction face detection processing module at the subsequent stage of the filter processing module 301, and executes processing by the filter processing module 301 on image data as compared with the above-described reference processing model. The point to do is the same. On the other hand, the face detection processing in the row direction face detection processing module is different.

  Specifically, the face detection processing module 302 in the reference processing model executes face detection processing in the raster scan direction as shown in FIG. In contrast, the column direction face detection processing module in the derivation processing model executes processing in the direction as shown in FIG. Therefore, for example, as shown in FIG. 17B, faces are detected in four places in the image, but the addresses of the four places are stored in the array data in the order shown in FIG. 17B. The It is assumed that the processing module attribute “(8) column direction scan” is set in advance by the creator for the column direction face detection processing module in the derived processing model. In the derivation processing model, it is assumed that test data is not associated and managed.

  Next, the test data conversion tool in this embodiment will be described. Similar to the first embodiment, the test data conversion tool in this embodiment is executed as part of graphic programming application software. It is assumed that the user has created the derivation process model described above with respect to the reference process model that has been created in advance.

  First, in the test data conversion tool, the process model difference detection unit 101 detects a process model difference between the reference process model and the derived process model specified by the user. Similarly, in this embodiment, a list of processing modules to be connected is created. FIGS. 18A and 18B show an example of a list of processing modules created from the reference processing model and the derived processing model.

  Here, as shown in FIG. 18A and FIG. 18B, both processing models have different processing module attributes, and the difference between the processing models is detected as a difference between the processing module lists. In the present embodiment, as shown in FIGS. 18A and 18B, a difference is detected as a processing module attribute.

  Subsequently, in the test data conversion tool, processing by the search unit 102 shown in FIG. 1 is executed. Similar to the first embodiment, the search unit 102 searches the dictionary means for a test data conversion processing method using the difference between the processing models as a search key. When a difference between processing models (processing module processing attributes) is input as a search key, a conversion processing model registered as ID-jkl is searched as a test data conversion processing method that matches the search key as shown in FIG. The

  Subsequently, in the test data conversion tool, processing by the conversion processing unit 103 shown in FIG. 1 is executed. The conversion processing unit 103 converts the test data of the reference processing model into test data corresponding to the derived processing model based on the searched conversion processing model.

  The ID-jkl conversion processing model searched as described above is registered as a processing model including an array order conversion processing module 2601 as shown in FIG. The array order conversion processing module 2601 receives array data corresponding to the output data of the reference processing model, and converts the order of the elements of the input array data into the order of the elements of the array data corresponding to the output data of the derivation processing model. To do. That is, the array order conversion processing module 2601 is designed to output correct data in the test data of the derivation processing model when correct data in the test data associated with the reference processing model is input.

  Specifically, the order held in the array data of the correct answer data is converted into an order corresponding to the scan direction in the derivation process model. As a result, in the present embodiment, the conversion processing unit 103 converts the correct answer data in the test data associated with the reference process model into correct answer data corresponding to the derived process model as shown in FIG.

  As described above, when there is a processing model (reference processing model) associated with the test data in advance by the test data conversion tool in the present embodiment, a new test data is created for the newly created processing model. Time and effort is reduced. That is, when a user newly creates a processing model (derivative processing model) having a difference between processing models registered in the dictionary based on the processing model, test data for the derived processing model is automatically generated. Is possible.

(Fifth embodiment)
The fifth embodiment of the present invention will be described below with reference to the drawings. The test data conversion tool in this embodiment is different in that an execution hardware environment (target platform) set in the processing module attribute is detected as a difference between the reference processing model and the derived processing model. Accordingly, the test data conversion process is also different. Hereinafter, in the present embodiment, only differences from the first embodiment will be described, and description of common parts will be omitted. Further, the system configuration and the functional configuration of the PC according to the present embodiment are the same as those in the first embodiment, and thus description thereof is omitted.

  The reference processing model in this embodiment includes a CPU face detection processing module at the subsequent stage of the filter processing module 301. As described above, the processing by the filter processing module 301 is executed on the image data as compared with the reference processing model described in the first embodiment, but the CPU face detection processing module is different.

  For the CPU face detection processing module in the reference processing model, “(9) serial processing CPU” of the execution hardware environment for executing processing is set. That is, the CPU face detection processing module is designed to operate on a single core CPU. The processing content is the same as that of the first embodiment, and the test data that is managed in association is also the same as that of the first embodiment, so that the description thereof is omitted.

  Next, the derivation process model will be described. The derivation processing model in this embodiment includes a GPU face detection processing module at the subsequent stage of the filter processing module 301. Compared with the reference processing model described above, the processing by the filter processing module 301 is the same as that for the image data, but the GPU face detection processing module is different.

  For the GPU face detection processing module in the derivation processing model, an execution hardware environment of “(10) parallel processing order indefinite GPU” is set in advance as a processing module attribute by the creator. Therefore, the GPU face detection processing module is designed to operate on the GPU. In the derivation processing model, it is assumed that test data is not associated and managed.

  Here, when face detection processing is executed in parallel on the GPU, it is often impossible to control the order of processing executed in parallel. Therefore, for example, as shown in FIG. 21, when a face is detected at four places in the image by the GPU face detection processing module according to this embodiment, the order in which the addresses at the four places are arranged in the array data is changed. It is not possible to grasp in advance.

  Next, the test data conversion tool in this embodiment will be described. Similar to the first embodiment, the test data conversion tool in this embodiment is executed as part of graphic programming application software. It is assumed that the user has created the derivation process model described above with respect to the reference process model that has been created in advance.

  First, in the test data conversion tool, the process model difference detection unit 101 detects a process model difference between the reference process model and the derived process model specified by the user. Similarly, in this embodiment, a list of processing modules to be connected is created. 22A and 22B show an example of a list of processing modules created from the reference processing model and the derivation processing model.

  Here, as shown in FIGS. 22A and 22B, both processing models have different processing module attributes, and the difference between the processing models is detected as a difference in the list of processing modules. Specifically, as shown in FIGS. 22A and 22B, (9) the serial processing CPU and (10) the parallel processing order indefinite GPU are detected as differences in the processing module attributes. .

  Subsequently, in the test data conversion tool, processing by the search unit 102 shown in FIG. 1 is executed. Similar to the first embodiment, the search unit 102 searches the dictionary means for a test data conversion processing method using the difference between the processing models as a search key. When a difference between processing models (processing module processing attributes) is input as a search key, a conversion processing model registered as ID-mno is searched as a test data conversion processing method matching the search key as shown in FIG. The

  Subsequently, in the test data conversion tool, processing by the conversion processing unit 103 shown in FIG. 1 is executed. The ID-mno conversion processing model searched as described above is registered as a processing model including the array order conversion processing module 3201 as shown in FIG. The array order conversion processing module 3201 receives array data corresponding to the output data of the reference processing model.

  Here, the array order conversion processing module 3201 is designed to execute a predetermined process when correct data in the test data associated with the reference processing model is input. Specifically, the order in which the correct answer data is stored in the array data is sorted and converted into the order of the processing direction when the face detection process is executed in the raster scan direction. In this embodiment, since the face detection processing direction of the CPU face detection processing module in the reference processing model matches the raster scan direction, there is no change in the order in which the correct answer data is stored in the array data.

  Further, in the present embodiment, when testing the derivation process model, the detected conversion process method is applied to the output data of the derivation process model to be compared with the correct data to perform the conversion process. Specifically, as shown in FIG. 24, the order in which the output data of the derivation processing model is held in the array data is sorted and converted into the order of the processing direction when the face detection process is executed in the raster scan direction.

  That is, in the present embodiment, the hardware environment for executing the face detection process in the derivation process model is a GPU, and the order held in the array data of the output data cannot be controlled. Therefore, the test data associated with the reference processing model can be diverted by sorting both the correct answer data and the output data of the derived processing model by a predetermined sorting method.

  As described above, when there is a processing model (reference processing model) associated with the test data in advance by the test data conversion tool in the present embodiment, a new test data is created for the newly created processing model. Time and effort is reduced. That is, when a user newly creates a processing model (derivative processing model) having a difference between processing models registered in the dictionary based on the processing model, test data for the derived processing model is automatically generated. Is possible.

(Sixth embodiment)
Hereinafter, a sixth embodiment of the present invention will be described with reference to the drawings.
FIG. 25 is a flowchart illustrating an example of a test data conversion processing procedure executed by the test data conversion tool according to the present embodiment. The processing of S3301, S3302, and S3304 corresponds to the processing performed by the processing model difference detection unit 101, the search unit 102, and the conversion processing unit 103 described in each of the above embodiments. Therefore, description of these processes is omitted.

  As shown in FIG. 25, in the test data conversion tool of this embodiment, a dictionary addition registration process step (S3305) is further added to the processing procedures described in the first to fifth embodiments. That is, in S3303, the search unit 102 determines whether the search key is registered in the dictionary unit as a result of the search. If the search key is registered in the dictionary unit, the process advances to step S3304 to perform the conversion process described in each embodiment described above.

  On the other hand, if it is determined in S3303 that the search key is not registered in the dictionary unit and the test data conversion processing method cannot be searched, the process advances to S3305. In step S <b> 3305, the search unit 102 newly designates a test data conversion processing model by user operation and registers it in the dictionary unit.

  Specifically, in the test data conversion tool, as shown in FIG. 26A, first, a message 3401 indicating that the test data corresponding to the derived processing model cannot be automatically converted is displayed on the screen. To do. On the other hand, the user creates a new conversion processing model for converting test data or selects an existing conversion processing model and sets the ID of the conversion processing model as shown in FIG. Enter to register additionally in association with the search key.

  For example, in the first to fifth embodiments, the conversion processing model described in each embodiment has been described as already registered in the dictionary unit, but when they are not registered in the dictionary unit, A case where the user newly registers in the dictionary means can be considered. In addition, a conversion processing model may be newly registered in association with a search key other than the case described in the first to fifth embodiments.

  As described above, in the test data conversion processing tool according to the present embodiment, a method for converting test data associated with a reference processing model using a difference between the reference processing model and a derived processing model as a search key is a dictionary value. Can be additionally registered. As a result, the user can increase the number of test data conversion processing methods registered in the dictionary at any time. Therefore, there is a case where the effort for newly creating test data is reduced as the number of registered data in the dictionary increases. It becomes possible to increase.

(Other embodiments)
The present invention supplies a program that realizes one or more functions of the above-described embodiments to a system or apparatus via a network or a storage medium, and one or more processors in a computer of the system or apparatus read and execute the program This process can be realized. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions.

101 processing model difference detection unit 102 search unit 103 conversion processing unit

Claims (10)

An information processing apparatus that creates test data corresponding to a newly created second processing model using a first processing model in which test data is associated and managed with respect to a processing model composed of processing modules. There,
Detecting means for detecting a difference between the first processing model and the second processing model;
Search means for searching a dictionary corresponding to the difference detected by the detection means from a dictionary that holds the difference between the first processing model and the test data conversion processing method in association with each other;
An information processing apparatus comprising: conversion means for converting test data related to the first processing model into test data related to the second processing model using the dictionary searched by the search means.   The information processing apparatus according to claim 1, wherein the detection unit detects at least one difference among a topology, a processing module attribute, a target platform, and a data size input / output to / from the processing model. The detection means detects a difference regarding the attribute of the processing module related to the detection direction,
The conversion means converts the order of test data related to the first processing model into the order of test data related to the second processing model using the dictionary searched by the search means. Item 3. The information processing device according to Item 2.   The information processing apparatus according to claim 3, wherein the detection unit further detects a difference regarding the topology. The detection means detects a difference regarding the data size input and output to the processing model,
The said conversion means extracts the test data which concerns on a said 2nd process model from the test data which concerns on a said 1st process model using the dictionary searched by the said search means. Information processing device. The detecting means detects a difference regarding the attribute of the processing module related to the processing range,
The said conversion means extracts the test data which concerns on a said 2nd process model from the test data which concerns on a said 1st process model using the dictionary searched by the said search means. Information processing device. The detecting means detects a difference with respect to the target platform;
The conversion means converts the order of test data related to the first processing model into the order of test data related to the second processing model using the dictionary searched by the search means. Item 3. The information processing device according to Item 2.   8. The method according to claim 1, further comprising: a registration unit that additionally registers the dictionary when there is no dictionary corresponding to the difference detected by the detection unit as a result of the search by the search unit. The information processing apparatus according to item 1. An information processing method for creating test data corresponding to a newly created second processing model by using a first processing model in which test data is managed in association with a processing model constituted by processing modules. There,
A detecting step of detecting a difference between the first processing model and the second processing model;
A search step of searching a dictionary corresponding to the difference detected in the detection step from a dictionary that holds the difference between the first processing model and the test data conversion processing method in association with each other;
An information processing method comprising: a conversion step of converting test data related to the first processing model into test data related to the second processing model using the dictionary searched in the searching step. An information processing apparatus that creates test data corresponding to a newly created second processing model using a first processing model in which test data is managed in association with a processing model constituted by processing modules. A program for controlling,
A detecting step of detecting a difference between the first processing model and the second processing model;
A search step of searching a dictionary corresponding to the difference detected in the detection step from a dictionary that holds the difference between the first processing model and the test data conversion processing method in association with each other;
A program causing a computer to execute a conversion step of converting test data related to the first processing model into test data related to the second processing model using the dictionary searched in the search step.