JP2017129996A - Coordination recommendation device, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To present a coordination fitting to user's body shape.SOLUTION: A silhouette extraction part 130 extracts a silhouette area. An index calculation section 132 calculates indexes based on feature quantity of the silhouette. A coordination determination section 134 determines a coordination of a user based on the calculated indexes and reference values corresponding to the indexes.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a coordination recommendation device and a program, and more particularly to a coordination recommendation device and a program for recommending coordination to a user.

  Conventionally, techniques related to methods for recommending coordinates have been proposed. Non-Patent Document 1 records daily coordinates together with information such as the weather and meeting people, and recommends coordinates by referring to the past history according to weather forecasts, schedules, meeting people, etc. based on such information. Technology has been proposed.

  Non-Patent Document 2 proposes a technique in which the coordinates of the day are photographed and recorded when standing in front of a mirror, and the coordinates are recommended from the past user's history.

Satoshi Hamada, Kaori Kitamura, Keisuke Kambara, Koji Tsukada, Ichiro Shiio, “Asa1-coordinator: Supporting Fashion Coordination Using History Information” Human Interface Symposium 2009 Ayaka Sato, Keita Watanabe, Michiyasu Yasumura, “Fashion Coordination Support System using figure suGATALOG” Information Processing Society of Japan Interaction 2011

  However, Non-Patent Document 1 and Non-Patent Document 2 do not consider whether past history data is suitable for the user's body shape or more objectively more suitable. In fashion, viewpoints such as time and place, situation, combination of items, and preference for the atmosphere of coordination are important, but in addition, the viewpoint of coordination that fits the user's system is also important. Whether coordinates are in the user's body in the first place, make the area of interest of the body inconspicuous, judge whether the appearance is close to the ideal body, and recommend coordinates so that you can cover the body you care more about There was no system to do. Whether or not you can cover the figure you are interested in is not something that can be overcome simply by measuring the size and wearing something that fits the size. For example, wearing loose clothing will make the body look relatively thin. Coordination aimed at effect was necessary, and there was no system that could make such a recommendation.

  In general, an illusion or the like is used for coordination in order to cover a figure of interest. For example, in fashion magazines, if you are worried about a small figure, choose a small pattern, if you want to show your legs long, put a belt slightly above the waist and raise the center of gravity, etc. Coordination that introduces or distracts attention from a point of concern is introduced. However, even if ordinary people who are not fashion coordinators try to consider these techniques, there is a problem that it is difficult to apply them to actual coordination.

  The present invention has been made to solve the above-described problems, and an object thereof is to provide a coordination recommendation device and a program that can determine a coordination suitable for a user's body shape.

  In order to achieve the above object, a coordination recommendation device according to a first aspect of the present invention is a coordination recommendation device that presents a coordinate that matches a user's body shape, and is a silhouette region that represents the silhouette of the user from the coordinated image of the user A silhouette extracting unit for extracting the index, an index calculating unit for calculating each index obtained from a feature amount of the silhouette based on the silhouette region extracted by the silhouette extracting unit, and each of the calculated by the index calculating unit The coordinate determination part which determines the said user's coordination based on a parameter | index and each reference value with respect to each said parameter | index is comprised.

  Further, in the coordination recommendation device according to the first aspect, the coordination determination unit is configured based on a comparison result obtained by comparing each index calculated by the index calculation unit with each reference value for each index. You may make it acquire the advice regarding coordination predetermined with respect to the comparison result.

  In the coordination recommendation device according to the first aspect, the feature amount of the silhouette may include a salientity map of the silhouette area, a width of each part of the silhouette area, and a center of gravity of the silhouette area.

  In the coordination recommendation device according to the first aspect of the present invention, the coordinate recommendation device further includes a coordination determination database storing advice relating to coordination, which is predetermined for the comparison result, for each of the comparison results, You may make it acquire the advice regarding the said coordination with respect to the said comparison result from the said coordination determination database.

  A program according to the second invention is a program for causing a computer to function as each part of the coordination recommendation device according to the first invention.

  According to the coordinate recommendation device and the program of the present invention, a silhouette region is extracted, each index obtained from the silhouette feature amount is calculated, and based on each calculated index and each reference value for each index, By determining the user's coordinates, the effect that the coordinates suitable for the user's body shape can be determined is obtained.

It is a figure which shows the structure of the coordination recommendation system which concerns on embodiment of this invention. It is a block diagram which shows the structure of the coordination recommendation apparatus which concerns on embodiment of this invention. It is a figure which shows an example of an attitude | position when a user is image | photographed. It is a figure which shows an example of the division | segmentation of a silhouette area | region, and a detection. It is a figure which shows an example of each part of a silhouette area | region. It is a flowchart which shows the coordination recommendation process routine in the coordination recommendation apparatus which concerns on embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<Overview according to the embodiment of the present invention>

  The outline | summary which concerns on embodiment of this invention is demonstrated. For fashion, other colors, patterns, designs, and other combinations of silhouettes, patterns, etc. There seems to be a rule regarding. Then, it is considered that the rules can be digitized and presented as an easy-to-understand index for general users, so that it is possible to recommend a better coordination example suitable for the individual user according to the user's body shape. Therefore, in the system of the present embodiment, is it possible to express the fashion rules with indices so that they can be easily understood by the general public from the user's coordinated image, and to cover well the figure of interest based on those numbers? If it is determined, and if it cannot be covered well, coordination advice indicating how to correct it is presented, and advice on coordination such as recommending more suitable items is given.

<Configuration of Coordination Recommendation System According to Embodiment of the Present Invention>

  First, the configuration of the coordination recommendation system according to the embodiment of the present invention will be described. As shown in FIG. 1, the coordination recommendation system 1 according to the embodiment includes a client terminal 50, a network 80, and a coordination recommendation device 100.

  The client terminal 50 receives an input of a coordinated image from the user and transmits the coordinated image to the coordinate recommendation device 100 via the network 80. Further, the client terminal 50 displays the recommendation result of coordination received from the coordination recommendation device 100 via the network 80 on the display of the client terminal 50 or the like.

  The coordination recommendation device 100 determines the coordination of the coordinated image based on the coordinated image received from the client terminal 50 via the network 80, and provides advice regarding the coordination corresponding to the determination result via the network 80 to the client. It transmits to the terminal 50.

<Configuration of Coordination Recommendation Device According to Embodiment of the Present Invention>

  Next, the configuration of the coordination recommendation device according to the embodiment of the present invention will be described. As shown in FIG. 2, the coordination recommendation device 100 according to the first embodiment includes a CPU, a RAM, and a ROM that stores a program and various data for executing a coordination recommendation processing routine to be described later. Can be configured. The coordination recommendation device 100 is functionally configured to include an input unit 110, a calculation unit 120, and an output unit 190 as shown in FIG.

  The input unit 110 receives a user's coordinated image received from the client terminal 50 via the network 80. Since the coordinated image is an image reflected in a mirror and is a smartphone terminal, the image may be taken by the user himself, or may be taken by attaching a Web camera or the like to the PC. It is a photograph taken in the appearance of the coordination you want to judge. For example, the image is taken from the height of the user's line of sight in the posture shown in FIG. In addition, the input unit 110 receives an input of an option representing a figure that the user is interested in. The choices are “lower body volume”, “low height”, “second arm thickness”, and the like.

  The calculation unit 120 includes a silhouette extraction unit 130, an index calculation unit 132, a coordinate determination unit 134, a coordinate presentation unit 136, and a coordinate determination database 140.

  The silhouette extraction unit 130 extracts a silhouette region representing the user's silhouette from the coordinated image received by the input unit 110. The silhouette extraction unit 130 divides the extracted silhouette region into parts. In the case of fashion, for example, in the case of fashion, for example, tops and bottoms, and tops and inners and outers, other hats, mufflers, scarves, and belts can be detected. Using methods such as those in Reference Document 1 and Reference Document 2, the target region is cut out and parts are detected.

[Reference 1] Body Part Detection for Human Pose Estimation and Tracking Mun Wai Lee and Ram Nevatia IEEE Workshop on Motion and Video Computing (WMVC'07)

[Reference 2] Kota Yamaguchi, M Hadi Kiapour, Luis E Ortiz, Tamara L Berg, "Parsing Clothing in Fashion Photographs", CVPR 2012

  For example, as shown in 401 of FIG. 4, the parts are detected by dividing the silhouette region into two equal parts in the height direction and dividing the upper part into tops and the lower part into parts. In addition, for the detection of small parts, for example, an edge is detected from the image, the silhouette center part, for example, the silhouette area is divided into three equal parts in the height direction, and the horizontal edge whose intensity in the middle area is equal to or greater than the threshold May be a belt. Further, as shown in 402 of FIG. 4, face detection may be performed, and it may be determined that a stall has been applied if an edge whose edge strength is equal to or greater than a threshold value exists in the lower part of the face. The method described in Reference 3 may be used for face detection.

[Reference 3] Rainer Lienhart and Jochen Maydt, "An Extended Set of Haar-like Features for Rapid Object Detection", IEEE ICIP 2002, Vol. 1, pp. 900-903, Sep. 2002.

  The index calculation unit 132 performs the following first to sixth processes, and calculates each index obtained from the silhouette feature amount based on the silhouette region extracted by the silhouette extraction unit 130.

  Specifically, the feature amount of the silhouette that is the basis of the index is calculated in the first to fifth processes, and the index value of each index is calculated in the sixth process. The feature amount of the silhouette includes a salientity map of the silhouette area calculated in the first process, the center of gravity of the silhouette area calculated in the second process, the width of each part of the silhouette area calculated in the third process, These are the balance between the upper body and the lower body calculated by the above process, the size of the pattern calculated by the fifth process, and the like.

  First, in the first process, the index calculation unit 132 searches for an eye-catching region in the silhouette region from the extracted silhouette region. When the position of an item such as an accessory, a stall, or a belt is known by detection by the silhouette extraction unit 130, as shown in 403 in FIG. Determine as an area. Further, image processing is performed, a Saliency map is calculated, and a region having a high edge density or a region having a high Saliency value is determined as a region that attracts attention. Since fields such as fashion are conscious of human eyes, they are calculated by combining image features such as edges and colors. The Salientity map described in Reference Document 4 may be used for calculating the feature amount.

[Reference 4] L. Itti, C. Koch, E. Niebur, A Model of Saliency-Based Visual Attention for Rapid Scene Analysis, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 20, No. 11, pp. 1254 -1259, Nov 1998.

  Next, the index calculation unit 132 calculates the center of gravity of the silhouette region in the second process.

  The pixels in the silhouette region can be calculated with the same weight as the following equation (1) using the same weight.

In this embodiment, a certain part of the accessory detected when dividing into parts is weighted, for example, twice as much weight, and the center of gravity of the silhouette region is calculated using the value of the Saliency map as a weight. The weight in each pixel is set to f (x _i , y _i ), and the center of gravity of the silhouette region is calculated according to the following equation (2).

  When the belt is detected in the first process, the center of the belt may be set as the center of gravity.

  Next, in the third process, the index calculation unit 132 measures the width (number of pixels) of each part of the silhouette region including the width of each part on the image. For example, as shown in FIG. 5, the width of each part is a face width 501, a shoulder width 502, a waist width 503, a hip width 504, a width 505 at the most widened position with clothes, an arm width, an ankle width, etc. The width is the width of the wrist. You may specify the position which measures a width | variety manually, for example, the height which hits a shoulder, etc. When automatically specifying the height to measure the width, for example, the shoulder width is detected by detecting the face, taking the widest width within that area as the face width and calculating the width from the face area down to the face area. Let's assume that the width is twice or more than the face width. The hip width is a width at a height corresponding to half the height of the silhouette area. The waist width is the width at the belt position when a belt is detected, and the width at the elbow height when the arm is folded, or the narrowest width between the shoulder and hip. It may be width. A method may be used in which the height between the shoulder height and the hip height is divided into three equal parts, and the height position closer to the hip is designated as the waist. Record the widest width of the silhouette area.

  Here, when the silhouette area is divided into ankles, wrists, and thighs as parts, the widths at the positions of the ankles, wrists, and thighs may be measured. In addition, when the clothes pattern is a stripe, a check pattern, a flower pattern, etc., if the width of the line and the interval between the patterns can be measured, they may be measured.

  Next, the index calculation unit 132 calculates the balance between the upper body and the lower body in the fourth process. In the fourth process, first, after detecting a part, a boundary line between the upper body and the lower body is set. For example, the boundary between tops (jackets, T-shirts, etc.) and bottoms (pants, skirts, etc.), belt position, the narrowest part between the shoulders and hips, and the upper and lower boundaries on the design. And the boundary. In the case of one-piece designs with upper and lower switching lines, the switching position is calculated. If a plurality of values can be calculated, the highest position is detected as a boundary line between the upper body and the lower body. Next, each of the number of pixels in the region occupied by the upper body (tops) and the number of pixels in the region occupied by the lower body (bottoms) is calculated by the following equation (3). If an area that catches the eye of a person is detected, the area is calculated with weighting.

  Here, α represents the degree of attention in the pixel. For example, it is normally set to 1 and an area with accessories such as scarves and belts is set to 2. Alternatively, the value of the Saliency map may be α and the sum may be obtained. n represents the number of pixels of the upper body or the lower body.

  Next, in the fifth process, the index calculation unit 132 calculates the pattern size for a part that can calculate the pattern size in the silhouette region. For example, in the case of a stripe, the width of the stripe, and in the case of a pattern, the size of the pattern is calculated, and the ratio between the silhouette area and the size of the pattern is calculated.

  Finally, as the sixth process, the index calculation unit 132 calculates the index values of the following indices 1 to 4 using the feature amounts of the silhouette regions obtained in the first to fifth processes.

Index 1: Hip width / widest width index in silhouette area 2: Waist width / widest width in silhouette area index 3: Height of center of gravity index 4: Number of pixels in lower body / number of pixels in upper body

  In the present embodiment, the index values of the index 1 to index 4 are calculated. However, the present invention is not limited to this, and other indices may be calculated. For example, the width of the arm and the width of the cuff of the clothes may be calculated, and the index value of the width of the arm / the width of the cuff may be calculated. Further, when the pattern width is calculated, the pattern width / height may be calculated as an index.

  In addition, the index calculation unit 132 determines an eye-catching region using the feature amount of the silhouette region. For example, among the silhouette regions, a region having a high edge density or a region having a high value of Saliency is determined, and how much eye-catching is expressed by a gray value.

  The coordination determination database 140 stores advice on coordination. For advice, for example, a comment when the value of each index does not satisfy the standard value as follows is stored. When the index value 1> 1, the advice “Wear a wide skirt such as an A-line skirt or a jacket” is stored. In addition, when the index value is 4> 1, the advice “Let the attention to the upper body by putting a stall around the neck” is stored.

  The coordinate determination database 140 stores reference values for each index. Each reference value for each index is, for example, 1 for index 1 and a value less than 1 for index 1, 2 for reference 1 and a value less than 1 for index 3, For example, 1/2 is a reference value, a value higher than 1/2 is good, and an index of 4 is 1 as a reference value, and a value less than 1 is good. In addition to this, for example, when the thickness of the second arm is a concern, a reference value may be provided such that the smaller the index value of the arm width / cuff width, the “good”. In addition, it is preferable that the reference value stored in the coordination determination database 140 and the comment of advice are prepared in advance by a fashion coordination professional or the like. In addition, the coordinate determination database 140 may store a comment when the ratio of each width of the ideal body (for example, the ideal value of each index) is compared with the ratio of each width of the user. For example, if "ideal ratio (index value)> user ratio (index value)", "ideal ratio = user ratio", and "ideal ratio <user ratio" You may leave it. For the ideal body ratio, you may refer to numerical values that are generally described in magazines, etc., or collect coordinate images that are said to be ideal such as models, calculate the average of the numerical values of the indicators, and calculate the value It is good also as an ideal ratio.

  The coordinate determination unit 134 compares each index calculated by the index calculation unit 132 with each reference value for each index to determine the user's coordinate. Then, the coordination determination unit 134 obtains advice on coordination for the comparison result from the coordination determination database 140 based on the comparison result between each index and each reference value, and generates advice on coordination.

  For example, when “lower body volume” is input as an option in the input unit 110, an index relating to the option is selected, and a comment regarding advice relating to coordination is referred to with reference to a reference value for the index stored in the coordination determination database 140 Select.

  In this embodiment, when “lower body volume” is input, reference values corresponding to the index 1 and the index 4 are referred to, and it is determined whether the index 1 and the index 4 satisfy the reference value. Here, it is determined whether index 1 is larger or smaller than 1, and index 4 is larger or smaller than 1. When the index value of the index 1 and the index 4 is less than 1, an advice that the coordination is good is generated. When the index value of the index 1 or the index 4 is greater than 1, refer to the coordination determination database 140 Generate advice. For example, when both index 1 and index 4 are 1 or more, the coordinate judgment database 140 advises that index 1> reference value 1 “This coordination makes the lower body stand out. Wear a wide skirt or jacket. “Let's wear it”, get the advice “Let's focus on the upper body by putting a stall around the neck, etc.” when index 4> reference value 1, and the lower body stands out in this coordination. “Wear a wide skirt or jacket (index 1> 1). Put a stole around the neck to focus attention on the upper body (index 4> 1).” .

  Further, when “low height” is input, reference values corresponding to the index 3 and the index 4 are referred to and it is determined whether the index 3 and the index 4 satisfy the reference value. For example, when the standard of the index 3 is 1/2 or more and the standard of the index 4 is less than 1, a comment indicating that the coordination is satisfactory is generated because the standard value is satisfied, and when the standard value is not satisfied , Refer to the coordination judgment database 140, and advice when index 3 <reference value 1/2 "Let the tops be short. Let the tops have accessories. Let the belts be slightly above the waist." , When the index 4 is greater than the reference value 1, we get the advice "Let's focus on the upper body by putting a stall around the neck." Shorten the tops, bring the accessories to the tops, tighten the belt slightly above the waist (indicator 3). By, for example, sow, Try to go attention to the upper body (index 4). To generate the advice that ".

  In addition to the above, as an index, if the rule of color scheme can be converted into a numerical value, it may be used for determination, and the ratio of the area of cold color and warm color may be used as an index.

  The coordinate presentation unit 136 outputs the advice relating to the coordination generated by the coordination determination unit 134 and the image representing the eye-catching region of the silhouette region determined by the index calculation unit 132 to the output unit 190, and the network 80. The result is sent to the client terminal 50 as a coordinate recommendation result. At this time, in the output image, a part that is a target of coordination advice may be noted.

<Operation of the coordination recommendation device according to the embodiment of the present invention>

  Next, the operation of the coordination recommendation device 100 according to the embodiment of the present invention will be described. When the input unit 110 accepts a coordinate image and an option representing a body shape that the user is interested in, the coordinate recommendation device 100 executes a coordinate recommendation processing routine shown in FIG.

  First, in step S100, a silhouette region representing a user's silhouette is extracted from the coordinated image received by the input unit 110, and the extracted silhouette region is divided into parts.

  Next, in step S102, image processing is performed on the silhouette region extracted in step S100, a Saliency map is calculated, and an edge is extracted.

  In step S104, the center of gravity of the silhouette region is calculated for the silhouette region according to the above equation (2).

  In step S106, the width (number of pixels) of each part of the silhouette area including the width of each part is measured on the image.

  In step S108, the upper body, lower body, and boundary lines are detected, and the number of pixels in the region occupied by the upper body and the number of pixels in the region occupied by the lower body are calculated by the above equation (3), so that the upper body and lower body Calculate the balance.

  In step S110, the size of the pattern is calculated for a part that can calculate the size of the pattern in the silhouette region.

  In step S112, the index values of the indexes 1 to 4 are calculated using the feature amounts of the silhouette regions calculated in steps S102 to S110.

  Next, in step S113, based on the Saliency map calculated in step S102 and the edge extraction result, an eye-catching region in the silhouette region is determined.

  In step S114, an index relating to the option received by the input unit 110 is selected, and for each selected index, the index calculated in step S112 is compared with each reference value for each index to determine the user's coordination. To do. If all the selected indices satisfy the reference value, the process proceeds to step S116, and if any of the indices does not satisfy the reference value, the process proceeds to step S118.

  In step S116, advice indicating that the coordination is good is generated.

  In step S118, based on the comparison result between each index and each reference value in step S114, advice on coordination for the comparison result is acquired from the coordination determination database 140, and advice on coordination is generated.

  In step S120, the advice regarding the coordination generated in step S116 or step S118 and the image representing the eye-catching region of the silhouette region obtained in step S113 are output to the output unit 190, via the network 80, It transmits to the client terminal 50 as a coordination recommendation result.

  As described above, according to the coordinate recommendation device according to the embodiment of the present invention, the silhouette region is extracted, each index obtained from the silhouette feature amount is calculated, each calculated index, and each index is calculated. By comparing each reference value to determine the user's coordinates, and obtaining the coordination advice for the comparison result from the coordination determination database 140 and generating the coordination advice, the coordination according to the user's body shape Can be recommended.

  In addition, according to the above-described embodiment, an index for determining whether or not a body figure to be covered can be determined, and a reference value for determining coordination based on the index value, and an input coordinate image Even if you don't have specialized knowledge about fashion, you can understand the tricks of coordination in an easy-to-understand manner by recommending clothes that can cover the body shape by using the coordinate determination database 140 It is possible to select clothes suitable for each user.

  The present invention is not limited to the above-described embodiment, and various modifications and applications can be made without departing from the gist of the present invention.

50 Client terminal 80 Network 100 Coordinate recommendation device 110 Input unit 120 Calculation unit 130 Silhouette extraction unit 132 Index calculation unit 134 Coordinate determination unit 136 Coordinate presentation unit 140 Coordinate determination database 190 Output unit

Claims (5)

A coordination recommendation device that presents coordination that matches the user's body shape,
A silhouette extraction unit that extracts a silhouette region representing the user's silhouette from the user's coordinated image;
Based on the silhouette region extracted by the silhouette extraction unit, an index calculation unit that calculates each index calculated from the feature amount of the silhouette;
A coordinate determination unit that determines the user's coordination based on each index calculated by the index calculation unit and each reference value for each index;
Coordinate recommendation device including   The coordinate determination unit is configured to give advice regarding coordination based on a comparison result obtained by comparing each index calculated by the index calculation unit and each reference value for each index. The coordinate recommendation device according to claim 1 to be acquired.   The coordinate recommendation device according to claim 1, wherein the feature amount of the silhouette includes a salientity map of the silhouette area, a width of each part of the silhouette area, and a center of gravity of the silhouette area. For each of the comparison results, further comprising a coordination determination database storing advice relating to coordination predetermined for the comparison results;
The coordination recommendation device according to claim 2, wherein the coordination determination unit acquires advice on the coordination with respect to the comparison result from the coordination determination database.   The program for functioning a computer as each part of the coordination recommendation apparatus of any one of Claims 1-4.