JP2000285151A - Design evaluating device and design preparing device and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To speedily evaluate a design. SOLUTION: Concerning this design evaluating device, the design prepared by a user is inputted from an input part 10. An element required for evaluating the design is extracted from the input design by an analytic part 12 and evaluated by an evaluation part 14. The evaluation part 14 is composed of a neural network or data base learnt based on the evaluation value of a teacher designer to the sample design, and the evaluation value of the input design is calculated and outputted to an output part 16. The evaluation value of the color balance or form balance of the design can include a function value as well.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to automatic evaluation or improvement of a design.

[0002]

2. Description of the Related Art There is a subjective aspect in the appearance design evaluation of a product or building, and this subjective evaluation has an important influence on a user's evaluation of a product or the like. For example, even if a car has excellent mechanical performance such as an engine and undercarriage, it is not often sold when an external design is not accepted by a user. Elements of the subjective evaluation include those that can be expressed by "look" such as harmony of color, shape, and spatial arrangement, and sensual elements such as gentleness, sharpness, and cuteness obtained from the subject. Designers who design products and the like are required to create products that are preferred by as many people as possible, and also want to efficiently obtain subjective evaluations and create better designs.

[0003]

However, in the past, how individual designers improve their qualities, that is, how to create a design that is acceptable to the user, has been determined by the created design. Is to show others to others and wait for others' evaluation. Further, for a manufacturer that manufactures products, etc., searching for user preferences by trial and error is a very inefficient operation, and has a problem that costs and time are required.

The present invention has been made in view of the above-mentioned problems of the prior art, and has as its object the purpose of automatically evaluating a design created by a designer or the like in a short time, and further automatically improving the design so that the designer or the like can use it. An object of the present invention is to provide an apparatus and a method capable of efficiently creating a design by providing feedback.

[0005]

In order to achieve the above object, the present invention provides an input means for inputting a design to be evaluated, a storage means for storing a design and an evaluation of the design as a database, and the database using the database. Evaluating means for evaluating and outputting the design to be evaluated. The evaluation values of generally accepted designers (teacher designers) are compiled into a database,
By comparing and contrasting the data in this database with the design input by the user, an evaluation value of the input design can be quickly obtained.

Here, there is further provided type input means for inputting the type of the design to be evaluated, wherein the database is provided for each design type, and the evaluation means corresponds to the type input to the type input means. It is also preferable to evaluate the design to be evaluated using a database that performs the evaluation. By creating a database for each type of design,
Evaluation corresponding to the type of design is possible.

Further, the present invention provides an input means for inputting a target design, a storage means for storing an improved design for the design as a database, and an improving means for automatically improving and outputting the target design using the database. It is characterized by having.

[0008] An improved design for the input design can be automatically obtained by compiling a database of the improved design of the design by the teacher designer and comparing and comparing the design input by the user with the design stored in the database.

Also in this case, the apparatus further comprises a type input means for inputting the type of the target design, wherein the database is provided for each design type, and the improving means is adapted to the type input to the type input means. Preferably, the target design is improved using a corresponding database.

[0010] Further, there is provided an improved item input means for inputting the improved items by the improved means, and the improved means can be improved based on the improved items. As a result, it is possible to obtain a design having a degree of improvement according to the user's preference. The improved designs in the database may be stored in association with the improved items.

Further, the present invention provides an input means for inputting a design to be evaluated, a processing means for learning the evaluation of the design using teacher data, and an evaluation and output of the design to be evaluated using the processing means. And evaluation means. By learning using the evaluation value of the teacher designer as the teacher data, the evaluation value of the newly input design can be quickly obtained.

Here, it is preferable to use, for example, a known neural network as the processing means. The neural network is composed of an input layer, a hidden layer, and an output layer, and each layer includes a plurality of neurons. A neuron is a multi-input / 1-output nonlinear signal processing element, and corresponds to a human nerve cell. The neurons in each layer are connected with a degree of connection called weight, and the network is adjusted by adaptively adjusting the weight so that a desired result is obtained from the output layer when a predetermined input value is given to the input layer. You can "learn". Applying this principle, a sample design is given to the input layer, and the weight between neurons is adjusted so that the evaluation value of the teacher designer for this sample design can be obtained from the output layer. An evaluation value similar to the evaluation value of the designer can be obtained. In this case, the teacher data is a combination of the sample design and the evaluation value.

Further, the present invention provides an input means for inputting a target design, a processing means for learning improvement of the design using teacher data, and an improvement means for improving and outputting the target design using the processing means. And characterized in that: By learning the improved design of the teacher designer using the teacher data, it is possible to quickly obtain an improved design corresponding to the newly input design.

The processing means can also be constituted by, for example, a neural network. In this case, a sample design is provided to an input layer, and learning is performed by adjusting weights between neurons in each layer so that an improved design by the teacher designer for the sample design is output from the output layer. The teacher data is a combination of the sample design and the improved design.

The present invention also provides a method for evaluating a design. The method includes the steps of inputting a design to be evaluated, and evaluating the design to be evaluated by comparing the existing design with the design to be evaluated using a database storing evaluations of the existing design.

The present invention also provides a method for creating a design. The method includes the steps of inputting a target design and automatically improving the target design by comparing the existing design with the target design using a database that stores an improved design for an existing design.

The present invention also provides another method for evaluating a design. In this method, a design to be evaluated is input to an input layer of a neural network that has been trained using teacher data on design evaluation, and an evaluation value of the design to be evaluated is output from an output layer.

The present invention also provides another method for creating a design. In this method, a target design and improvements are input to an input layer of a neural network that has been trained using teacher data on design improvement, and a design in which the improvements are improved is output from an output layer.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings, taking a car design and a clothing design as examples.

<Automotive Design> FIG. 1 is a block diagram showing the configuration of this embodiment. This device has an input unit 1
0, an analysis unit 12, an evaluation unit 14, and an output unit 16.

The input unit 10 further includes a user-created image input unit 10a, a vehicle type input unit 10b, and a functional value input unit 10c. The user created image input unit 10a inputs a design created by a user (designer) as a paper medium or as electronic data. When a paper medium is used, it is preferable to convert the data into digital data using a scanner or the like and input the digital data. Also, the vehicle type input unit 10
“b” specifies and inputs the type of automobile in the user-created image. Examples of the type of vehicle include a sports car, a sedan, a box van, and a wagon. The sensory value input unit 10c
This is used when the user wants an improved design (improved design), not limited to the evaluation of the design created by the user, and inputs the desired improvement degree for the sensory items such as gentleness, softness, and brightness. . For example, if the user wants to increase the kindness received from the whole design,
If the user inputs 2 and wants to slightly reduce the sporty feeling received from the entire design, he or she may input -1. When +2 is input for the gentleness, it means that the gentleness is input as the improvement item, and +2 is input as the improvement degree. Although the sensory items are optional, the figure illustrates “gentleness”, “softness”, “brightness”, “volume feeling”, and “sporty feeling”. The width of the sensory value, that is, the degree of improvement can be arbitrarily set. For example, the maximum value is +5, the minimum value is -5, and the status quo is 0.
It can also be. If the user only wants to evaluate, there is no need to enter these items. Input unit 10
Can be constituted by a keyboard, a mouse, a scanner, and a reading device for various recording media such as an FD, a CD, and a DVD.

The analysis unit 12 extracts a target component to be evaluated from the image created by the user, and is executed based on the vehicle type input by the user. Specific components include the entire body, the front part, the rear part, the front wheel part, and the like, and the analysis unit 12 extracts these by appropriately using them according to the vehicle type. The extracted components are
Further, it is digitized as a feature amount and supplied to the evaluation unit 14. The analysis unit 12 can be configured by a CPU of a computer. It is also possible to adopt a configuration in which a user inputs components from the input unit 10 instead of extracting components by the analysis unit 12. Therefore, the user may input in the case of design and vehicle type, the case of design and vehicle type and constituent elements, the case of design and vehicle type and desired sensory value, the case of design, vehicle type, component and sensory value, and the like.

The evaluation unit 14 evaluates the user-created image using the feature amount input from the analysis unit 12 (or the input unit 10), and outputs an evaluation value to the output unit 16 for each predetermined item. Although the evaluation items can be set arbitrarily, the color balance and the shape balance are considered as the evaluation items because the color balance and the shape balance are considered to be the main factors of the design. In addition, in addition to these, a sensory value such as gentleness can also be cited as an evaluation item. Further, when the user inputs a desired sensory value from the sensory value input unit 10c, the image created by the user is improved (improved) based on the sensory value and output to the output unit 16. As a specific configuration for evaluation and improvement, there is a circuit configuration having a learning function such as a neural network or a database.

When a neural network is used, learning is performed using evaluation values of designers (referred to as teacher designers) that have been highly evaluated by general designers. That is, a plurality of sample designs are input from the input layer, and teacher data is provided so that the evaluation value of the teacher designer is output from the output layer, and the degree of connection (weight) between neurons is adaptively changed. For the learning, for example, a known back propagation method can be used. Further, a plurality of teacher designers may exist. By learning using many sample designs and evaluation values for the sample designs, it is possible to output the evaluation values for an arbitrary design after learning. Regarding the design improvement (improvement), a plurality of sample designs are input from the input layer, and teacher data is provided so that the improved design of the teacher designer is output from the output layer, and the weight between neurons is changed. When the weight is changed, the learning is performed by changing the sensory value stepwise. Specifically, when the gentleness is set to +1, the weight is adjusted so that the improved design of the teacher designer can be obtained by slightly giving the gentleness to the sample design (the degree is based on the subjectivity of the teacher designer). When the gentleness is set to +2, the weight is adjusted so as to obtain an improved design of the teacher designer which gives the gentleness to the sample design. Not only gentleness, but also sensory values are subjective and may vary between teachers and designers, so it is also preferable to use teacher data from multiple teachers and designers to learn to obtain an intermediate design between them. It is. In order to obtain an intermediate design, the evaluation value of the improved design by the teacher designer may be associated with the improved design.

When a database is used, similarly, evaluation values of teacher designers for many sample designs are associated with each other and converted into data. When converting the data, it is preferable to convert the data for each vehicle type. A database for a sports car, a database for a sedan, a database for a wagon, and the like. When an unknown design is input, the unknown design vehicle type and the sample design in the same type of database are sequentially compared with the input design, and an evaluation value corresponding to the closest sample design, or a plurality of evaluation values close to the input design, By calculating the average of the evaluation values of the sample designs, it is possible to output the evaluation values for an arbitrary design. The same applies to the design improvement. The improved design of the teacher designer and the evaluation value of the improved design are associated with the sample design and converted into data, and the improved design having the sensory value closest to the sensory value input by the user is output. . Of course, when a database is used, the number of samples has an effect on the accuracy of the evaluation value or the quality of the improvement, so it is necessary to make as many sample designs as possible into a database.

The output unit 16 outputs an evaluation value from the evaluation unit 14,
Alternatively, if the user desires, an improved design (improved design) is output. The output unit 16 can be constituted by a display or a printer.

The analyzing section 12 and the evaluating section 14 in FIG. 1 can be constituted by a computer having a CPU and a memory.

FIG. 2 shows a processing flowchart in the analysis unit 12 when the user inputs a user-created design and a vehicle type. First, when the user inputs the name of the vehicle type and the created design from the input unit 10 (S10
1), an element characterizing the vehicle type is selected (S102).
For example, when the vehicle type is a sedan, the overall shape of the body, the rear portion, and the like are included. Next, the user-created design is divided into element regions characterizing the impression of the vehicle (S103), and design data serving as constituent elements is extracted (S104).
Then, the extracted design data is digitized as a feature value (S105).

FIG. 3 shows an example of digitizing the design data extracted in S105 as feature values. As described above, the appearance design impression is determined by the color and shape. Therefore, data obtained from the design for the color and shape are used as feature parameters, and the feature value of each element is calculated for each feature parameter. For example, when the vehicle type is a sedan, if the components that characterize the impression of the sedan are the entire body, the front part, the front wheel part,... Since the shape characteristic parameters include the center of gravity, the aspect ratio, the curvature, and the like, each is calculated from the design data, and the color characteristic parameter is the color center (color center of gravity;
UV, HSV) can be used, and there are color distributions and the like.

FIGS. 4 and 5 show examples of data given to the neural network as teacher data or data stored in the database. In both figures, (a) is the sample design, (b) is the evaluation value of the teacher designer for the sample design of (a),
(C) is an improved (improved) design by the teacher designer for the sample design of (a), and (d) is an evaluation value of (c). FIG. 4 (c) is an improved design in which the sporty feeling is mainly increased with respect to the sample design (a), and FIG. 5 (c) is a case where the softness is mainly increased with respect to the sample design (a). It is an improved design. By using a large number of these teacher data, the evaluation unit 14
The evaluation value of the design created by the user, and further, the improved design can be output using the element feature amounts obtained in the above processing.

<Clothing Design> FIG. 6 is a block diagram showing the configuration of the present embodiment. As in FIG. 1, it includes an input unit 10, an analysis unit 12, an evaluation unit 14, and an output unit 16.

The input unit 10 inputs a user-created design and a type of clothing. The types of clothing are, for example, pants, shirts, dresses, sweaters, suits, and the like. When the user desires an improved design, a sensory value can be input. The sensory items include, for example, gentleness, gorgeousness, brightness, volume, and calmness.
The maximum value of the sensory value is +5, as in the case of car design.
The minimum value can be set to -5, and the status quo can be set to 0.

The analysis unit 12 extracts the target constituent elements from the design created by the user according to the type of the clothing, and quantifies them as characteristic values. For example, in the case of pants, the constituent elements are the entirety, the hem, the hips, and the like. The shape characteristic parameters include the thickness of the leg and the degree of spread of the hem, and the color characteristic parameters include the color center and the color distribution. Therefore, these are calculated based on the design data.

The evaluation unit 14 is composed of a neural network or a database as in the case of a car design, and learns evaluation values by a teacher designer as teacher data, or converts the evaluation value into an arbitrary design when an arbitrary design is input. An evaluation value or an improved design is created and output to the output unit 16. When a neural network is used, a plurality of sample clothing designs are input from the input layer, and teacher data is provided so that the evaluation value of the teacher designer is output from the output layer, and the degree of connection (weight) between neurons is given.
Is adaptively changed. The sample clothing design and the evaluation value of the teacher designer for the sample clothing design are prepared for each type of clothing to train the neurons. Regarding design improvement (improvement), multiple sample designs are input from the input layer for each type of clothing, and teacher data is given from the output layer so that the improved design of the teacher designer is output, and the weight between neurons is changed. Let me do it.

When a database is used, similarly, evaluation values of teacher designers for many sample designs are associated with each other and converted into data. A database is constructed for each type of clothing. If an unknown design is entered, the unknown design is compared with the sample design in the same type of database and the input design in order, and the evaluation value corresponding to the closest sample design or a plurality of sample designs close to the input design By calculating the average value of the evaluation values, it is possible to output the evaluation value for an arbitrary design. The same applies to the design improvement, and data of the sample design is made in association with the teacher designer's improved design.

FIGS. 7 and 8 exemplify evaluation values of a teacher designer given to a neural network or a database as teacher data. In both figures, (a) is the sample design, (b) is the evaluation value of the teacher designer for the sample design of (a), (c) is the design of the teacher designer that is an improvement of the sample design of (a),
(D) is the evaluation value of (c). The difference between FIG. 7 and FIG.
The difference between the sample design and the improved design is that FIG. 7 shows an improved design in which gentleness is increased, whereas FIG. 8 shows an improved design in which mainly gorgeousness is increased. The learning unit 1 uses the sample designs and their evaluation values shown in FIGS.
4 can be configured.

Although the embodiments of the present invention have been described above, the present invention is not limited to automobile designs and clothing designs,
It can be applied to any products and buildings.

[0038]

According to the present invention, a design created by a designer or the like can be automatically evaluated in a short time, and can be automatically improved and presented to the designer, so that the designer can efficiently create the design. be able to.

[Brief description of the drawings]

FIG. 1 is a configuration block diagram of an embodiment of the present invention.

FIG. 2 is a processing flowchart of an analysis unit in FIG. 1;

FIG. 3 is an explanatory diagram of calculating a feature value.

FIG. 4 is a diagram (part 1) illustrating an example of evaluation values of a sample design in the case of an automobile design.

FIG. 5 is a diagram (part 2) illustrating an example of evaluation values of a sample design in the case of an automobile design.

FIG. 6 is a configuration block diagram of another embodiment of the present invention.

FIG. 7 is a diagram (part 1) illustrating an example of evaluation values of a sample design in the case of clothing design;

FIG. 8 is a diagram (part 2) illustrating an example of evaluation values of a sample design in the case of a clothing design.

[Explanation of symbols]

10 input units, 12 analysis units, 14 evaluation units, 16 output units.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Yoshiyuki Hatakeyama, Inventor 1 Toyota Town, Toyota City, Aichi Prefecture Inside Toyota Motor Corporation (72) Inventor Yoshiko Hara 1 Toyota Town, Toyota City, Aichi Prefecture Toyota Motor Corporation F Terms (reference) 5B046 AA04 AA10 DA02 FA04 GA01 HA05 KA05 5B050 BA18 CA07 EA13 FA02 FA13 5L096 BA12 FA60 FA68 FA70 9A001 BB06 HH01 HH05 HH06 HH23 JJ01 KK09 KK38 KK40

Claims (16)

[Claims] 1. An input means for inputting an evaluation target design, a storage means for storing a design and an evaluation of the design as a database, and an evaluation means for evaluating and outputting the evaluation target design using the database. A design evaluation device characterized by that: 2. The apparatus according to claim 1, further comprising a type input unit for inputting a type of the design to be evaluated, wherein the database is provided for each design type, and the evaluation unit is a type input unit. A design evaluation device for evaluating the design to be evaluated using a database corresponding to the type input to the design evaluation device. 3. An input means for inputting a target design, a storage means for storing an improved design for the design as a database, and an improving means for automatically improving and outputting the target design using the database. Characteristic design creation device. 4. The apparatus according to claim 3, further comprising a type input unit for inputting a type of the target design, wherein the database is provided for each design type, and the improving unit is provided for the type input unit. A design creation apparatus, wherein the target design is improved using a database corresponding to the input type. 5. The apparatus according to claim 3, further comprising an improvement item input unit for inputting an improvement item in the improvement unit, wherein the improvement unit performs improvement based on the improvement item. A design creation device characterized by performing. 6. An input means for inputting a design to be evaluated, a processing means for learning the evaluation of the design using teacher data, an evaluation means for evaluating and outputting the design to be evaluated using the processing means, A design evaluation device comprising: 7. The design evaluation apparatus according to claim 6, wherein said processing means is constituted by a neural network. 8. The design evaluation device according to claim 6, further comprising a type input unit for inputting a type of the design to be evaluated. 9. An input means for inputting a target design, a processing means for learning improvement of the design using teacher data, and an improving means for improving and outputting the target design using the processing means. A design creation device characterized by that: 10. An apparatus according to claim 9, wherein said processing means is constituted by a neural network. 11. The design creation apparatus according to claim 9, further comprising a type input unit for inputting a type of the target design. 12. The design creation apparatus according to claim 9, further comprising an improvement item input unit for inputting an improvement item in said improvement unit. 13. A method for evaluating a design, comprising the steps of: inputting a design to be evaluated; and using a database storing evaluations of an existing design, comparing the existing design with the design to be evaluated. And a step of evaluating the design. 14. A method for creating a design, comprising: inputting a target design; and using a database storing an improved design for the existing design, comparing the existing design with the target design, and comparing the target design with the target design. A step of automatically improving; and a method for creating a design. 15. A method for evaluating a design, comprising: inputting a design to be evaluated to an input layer of a neural network trained using teacher data on the evaluation of the design; and evaluating an evaluation value of the design to be evaluated from an output layer. A design evaluation method characterized by outputting 16. A method for creating a design, comprising: inputting a target design and an improvement to an input layer of a neural network trained using teacher data on design improvement; and improving the improvement from an output layer. A design creation method characterized by outputting a designed design.