WO2020156306A1 - 服饰搭配信息的处理方法、数据对象处理方法、系统及设备 - Google Patents
服饰搭配信息的处理方法、数据对象处理方法、系统及设备 Download PDFInfo
- Publication number
- WO2020156306A1 WO2020156306A1 PCT/CN2020/073140 CN2020073140W WO2020156306A1 WO 2020156306 A1 WO2020156306 A1 WO 2020156306A1 CN 2020073140 W CN2020073140 W CN 2020073140W WO 2020156306 A1 WO2020156306 A1 WO 2020156306A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- data object
- combination
- type
- rule
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Recommending goods or services
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
Definitions
- the present invention relates to the field of computer technology, and in particular to a processing method, data object processing method, system and equipment for clothing matching information.
- the common logic for recommending products on e-commerce platforms is a recommendation technology based on user behavior.
- Most of the results recommended to users are products that belong to the same category as the products that have been purchased, collected or viewed. For example, if users are interested in dresses, they often browse and click on dresses on the platform, and a variety of similar dresses will be displayed in the recommended traffic window. This method of recommendation will result in repeated recommendations. The user has clearly purchased the product, but the system repeatedly recommends similar products, which is likely to cause confusion for the user and a poor user experience.
- the various embodiments of the present application provide a clothing matching information processing method, a data object processing method, a system, and a device that solve or partially solve the above problems.
- a method for processing clothing matching information includes:
- a data object combination that meets the matching rules is selected as a clothing matching example that can be referenced by the user.
- a data object processing method includes:
- the at least one second data object is matched with the first data object, the at least one second data object is provided to the user.
- a method for processing clothing matching information includes:
- the clothing matching instance is a combination of data objects that meet the matching rules selected from a combination of multiple data objects; the combination of multiple data objects is a combination of at least one type that is different from the type of the first data object. Is generated by combining the second data object extracted from the data object set with the first data object.
- a method for processing clothing matching information includes:
- the data object combination that meets the matching rule among the multiple data object combinations is fed back to the client as a clothing matching instance.
- a system for processing clothing matching information includes:
- the client is configured to send request information corresponding to the specified operation to the server in response to the specified operation performed by the user on the first data object; receive the clothing matching instance that the server feedbacks on the first data object; The clothing matching instance is provided to the user;
- the server is used to obtain the request information sent by the client for the first data object; from a set of at least one type of data object that is different from the type of the first data object, extract the information used to communicate with the first data object
- the combined second data object is used to generate a plurality of data object combinations; the data object combination that meets the matching rule among the plurality of data object combinations is fed back to the client as a clothing matching instance.
- an electronic device includes a memory and a processor; among them,
- the memory is used to store programs
- the processor is coupled with the memory, and is configured to execute the program stored in the memory for:
- a data object combination that meets the matching rules is selected as a clothing matching example that can be referenced by the user.
- an electronic device in another embodiment, includes a memory and a processor; among them,
- the memory is used to store programs
- the processor is coupled with the memory, and is configured to execute the program stored in the memory for:
- the at least one second data object is matched with the first data object, the at least one second data object is provided to the user.
- a client device in an embodiment of the present application, includes a memory and a processor; wherein,
- the memory is used to store programs
- the processor is coupled with the memory, and is configured to execute the program stored in the memory for:
- the clothing matching instance is a combination of data objects that meet the matching rules selected from a combination of multiple data objects; the combination of multiple data objects is a combination of at least one type that is different from the type of the first data object. Is generated by combining the second data object extracted from the data object set with the first data object.
- a server device includes a memory and a processor; wherein,
- the memory is used to store programs
- the processor is coupled with the memory, and is configured to execute the program stored in the memory for:
- the data object combination that meets the matching rule among the multiple data object combinations is fed back to the client as a clothing matching instance.
- the technical solution provided by the embodiment of the present application extracts a second data object to be combined with the first data object from a data object set of at least one type different from the type of the first data object to generate multiple data object combinations, and then Use matching rules instead of strict matching templates to select clothing matching examples from multiple data object combinations that can be referenced by users.
- the universality is strong, and the output is richer under the premise of ensuring matching quality
- the technical solutions provided by the embodiments of this application can produce multiple sets of matching suggestions for 99% of the products in the target product pool, and the satisfaction rate of manual evaluation reaches 80%.
- FIG. 1 is a schematic flowchart of a method for processing clothing matching information provided by an embodiment of the application
- FIG. 2 is a diagram of data object circulation in which a method of processing clothing matching information is described in principle by taking a dress as an example in an embodiment of the application;
- FIG. 3 is a schematic diagram of a combination rule provided by an embodiment of this application.
- FIG. 4 is a table representing the style matching relationship according to an embodiment of the application.
- FIG. 5 is a schematic flowchart of a data object processing method provided by another embodiment of the application.
- FIG. 6 is a schematic structural diagram of a clothing matching information processing system provided by an embodiment of the application.
- FIG. 7 is a schematic flowchart of a method for processing clothing matching information according to another embodiment of the application.
- FIG. 8 is a schematic flowchart of a method for processing clothing matching information provided by another embodiment of this application.
- FIG. 9 is a schematic structural diagram of an apparatus for processing clothing matching information provided by an embodiment of the application.
- FIG. 10 is a schematic structural diagram of a data object processing device provided by another embodiment of the application.
- FIG. 11 is a schematic structural diagram of a clothing matching information processing device provided by another embodiment of the application.
- FIG. 12 is a schematic structural diagram of a clothing matching information processing device provided by another embodiment of the application.
- FIG. 13 is a schematic structural diagram of an electronic device provided by an embodiment of the application.
- the first one is based on manual editing
- the quality is high, but the number of matching is very small.
- the quality is very high, but due to human efficiency, the daily output is usually in the order of hundreds of thousands and the quantity is very small. Platform products are in the hundreds of millions, so a large number of products cannot be matched manually. Collocations that have nothing to do with the user's own purchased products will not work for the user.
- Plans based on related purchases are of average quality and large in quantity.
- a matching sample is formed, and the model is trained to automatically produce matching.
- This method will encounter the problem of too much noise in the training samples, because usually the clothing products purchased by users within a period of time are not guaranteed to be purchased for the same wear, so the quality of the training samples is not high (or the samples are processed to match the matching quality The required cost is too high), resulting in limited model effects; this method is only suitable for related product recommendation.
- the third type is based on templates
- the scheme based on matching templates is of high quality, with a large quantity under the open product set and a small quantity under the limited product set.
- manual matching is used as a template, and a single product is used to find similar functions.
- One set of matching is expanded into multiple sets of similar matching. This method can effectively improve the quality of a certain combination. Number of matches.
- the (manual) matching templates that the platform can accumulate are usually in the order of less than one million, which can meet the needs of platform products that can easily be tens of millions or even over 100 million (the number of combinations that can be arranged and combined is almost infinite)
- the products and the matching forms that can be covered are still limited.
- a matching template has high requirements for the richness of the product.
- a 4-piece collocation template means that the products of the specific tops, bottoms, shoes and bags are required to meet similar conditions before they constitute a qualified collocation under the template. Therefore, especially when seeking to produce matching under a limited product set (such as a certain brand, a certain store), the limited product candidates are not enough to meet the requirement that all parts of a specific template are similar, resulting in failure of matching output.
- the template-based method can still produce limited combinations, and cannot cover the matching needs of all products on the platform.
- a set of collocation usually consists of 3 to 5 products.
- clothing, shoes, and bags usually cover more than 40 types, and there are as many as tens of millions of product candidates for each type. Therefore, the search space for a set of matching product candidates is huge, and the product permutations and combinations that can be used as combination candidates are inexhaustible.
- the coverage of the method should reach more than 95%.
- the method requires a higher pass rate while ensuring the coverage and richness of the output mix.
- the calculation time required for a single product to return 10 sets and more collocations is less than 200ms.
- FIG. 1 shows a schematic flowchart of a method for processing clothing matching information provided by an embodiment of the present application.
- the execution subject of the solution provided in this embodiment may be a data object processing device, which may be a hardware with embedded program integrated on the client or server, or an application software installed on the client or server , It may also be tool software embedded in the client or server operating system, which is not limited in the embodiment of the present application.
- the client can be any terminal device including a mobile phone, a tablet computer, a PDA (Personal Digital Assistant, personal digital assistant), a POS (Point of Sales, sales terminal), a car computer, etc.
- the server can be a server, a cloud, etc.; this embodiment does not specifically limit this.
- the method includes:
- the above-mentioned first data object and second data object are specific clothing products, such as dresses, coats, shoes, bags, and so on.
- the first data object and the second data object may be a product number or a product name for uniquely identifying a product.
- the above steps can be triggered to be executed after the user performs any of the following actions:
- the user sends a voice requesting to recommend an example of clothing matching for product A;
- the user adds product A to the shopping cart
- the user purchased commodity A; etc.
- the object of the aforementioned user behavior—commodity A, is the first data object in this embodiment.
- the server may execute the above steps after receiving the matching request sent by the client for the first data object.
- the type to which the first data object belongs is the type of the commodity, such as shoes, clothing, bags, accessories (such as necklaces, headwear, earrings, rings, etc.), and so on. Because of the different wearing parts, clothing is divided into: tops, bottoms, conjoined and so on. Siamese category can be divided into: dress category, jumpsuit category and so on. Similarly, accessories can be divided into necklaces, headwear, earrings, rings, etc. because of the different wearing parts.
- the first data object is clothing information
- the type of clothing information is dress. Accordingly, at least one type of data object set as follows can be obtained:
- Footwear data object set is a Footwear data object set, accessory data object set, upload data object set, bag data object set, download data object set, etc.
- the technical solution provided in this embodiment may further include the following steps:
- the sorting rule may specifically be: inverted by popularity points + product on-shelf time.
- the number of selected data objects can be manually set, for example, ranking in the top 100, 150, 200 or more, which is not specifically limited in this embodiment.
- the type of the data object (such as shoes, clothing, etc.) and the sorting parameters (such as popularity score, shelf time, etc.) can be recorded in the attributes of the data object. That is, taking the data object set of the first type that is different from the type of the first data object as an example, the process of obtaining the data object set of the first type is described, which specifically includes:
- a plurality of data objects in the first order are selected to form the first type of data object set.
- the attributes of the data object may include: type, style, color information, texture information, sales record, additional purchase rate, click rate, shelf time, suitable season, suitable occasion, suitable style, etc.
- Popularity points can be calculated from sales records, additional purchase rates and click-through rates.
- the second data object can be directly extracted from the data object set, and combined with the first data object to obtain multiple data object combinations.
- the first data object is a dress
- a shoe data object set is obtained in step 101.
- the shoe data object set contains 200 data objects.
- the 200 data objects contained in the data object set can be combined with the dress. Get 200 data object combinations.
- the second data object can be extracted from part or all of the data object sets, and combined with the first data object to obtain multiple data object combinations.
- the first data object is a dress
- step 101 obtains a data object set of shoes, a bag data object set, and a jacket data object set.
- one data object a as the second data object can be extracted from the shoe data object set, and one data object b as the second data object can be extracted from the bag data object set, and then the data object a and data object b Combine with the first data object to form a data object combination; alternatively, you can extract a data object c as the second data object from the shoe data object set, and extract a data object d as the second data object from the bag data object set , Extract a data object e as the second data object from the jacket data object set, and then combine the data object c, data object d, and data object e with the first data object into a data object combination; and so on.
- the matching rule may include one or more rules, which is not specifically limited in this embodiment.
- matching rules may include: rules for styles, rules for colors, rules for bright spots, and so on.
- the above collocation rules can be obtained based on expert knowledge.
- the style-specific rules are used to filter out unmatched data object combinations such as composition structure and seasonal temperature, such as the data object combination of "T-shirt” + “down pants”.
- the color rules and the bright spot rules can be used to filter out data object combinations that do not meet the aesthetic requirements; the aesthetic requirements can be expert or expert's default dressing rules, such as no more than 3 main colors and no more than 1 bright spot. and many more.
- the method provided in this embodiment further includes:
- the purpose of adding the above steps is to further optimize the click-through rate of the match.
- the popularity scores of the commodities contained in each clothing matching instance can be sales volume, click-through rate, additional purchase rate, etc., which can be adjusted according to different business goals
- the score of each clothing matching instance can be calculated.
- the average value of the popularity scores of other data objects other than the first data object in the included products is used as the final calculated score.
- the technical solution provided in this embodiment extracts a second data object to be combined with the first data object from a set of at least one type of data object that is different from the type of the first data object to generate multiple data object combinations, and then use Matching rules, rather than strict matching templates, select clothing matching examples from multiple data object combinations that can be referenced by users. It is universally applicable and produces more rich combinations under the premise of ensuring matching quality. ; It can not only meet the actual matching needs of users, but also help to improve the commodity conversion rate of the e-commerce platform; in addition, through actual results, the technical solutions provided by the embodiments of this application can produce multiple sets of products for 99% of the target commodity pool Collocation suggestion, and the satisfaction rate of manual evaluation reaches 80%.
- step 102 "extract a second data object to be combined with the first data object from the set of at least one type of data object, To generate multiple data object combinations" the following steps can be used to achieve:
- a data object is selected from a part or all of the data object set of the at least one type of data object as the second data object and the first data object is combined into Data object combination; until the number of different data object combinations generated meets the preset number requirements.
- At least one combination rule may be obtained based on the type of the first data object.
- the obtained at least one combination rule may include: combination rules with shoes and bags 1. combination with jackets, shoes and bags Rule 2, Combination rule 3 with pants, shoes and bag combination, Combination rule 4 with coat, pants, shoes and bag combination, etc.
- the symbol in Figure 3 Characterization is optional.
- the obtained at least one combination rule may include: combination rule 1'with pants, shoes, and bags, and combination with skirts, shoes, and bags.
- the combination rules corresponding to each type can be set in advance. In this way, when a user makes a purchase, view, or application matching request for a certain type of data object, he can directly retrieve the combination rule corresponding to the type of the data object.
- step 1022 will be described by taking the at least one combination rule including: a first combination rule and a second combination rule as an example.
- a first combination rule and a second combination rule are only two types of data object sets or more than two types of data object sets require two or more combination rules. That is, when at least two types of data object sets are acquired in step 101, the foregoing step 1022 may specifically be:
- one data object is selected as the second data object and all the data object sets in the data object set of at least two types.
- the first data object is combined into a data object combination; until the number of different data object combinations generated meets the preset number requirement.
- this step 1022' can be implemented using the following steps:
- N3 1 for data object set 3
- N4 1 for data object set 4
- N1 1 for data object set 1.
- the preset number may be 100, 200, ... or more, which is not specifically limited in this embodiment.
- the data object set specified by combination rule 1 includes: data object set 3 and data object set 4; the data object set specified by combination rule 2 includes: data object set 1, data object set 3, and data object set 4; combination The data object set specified by rule 3 includes: data object set 2, data object set 3, and data object set 4; the data object set specified by combination rule 4 includes: data object set 1, data object set 2, data object set 3, and data Object set 4.
- the data object set specified by the combination rule can be one or more; among the multiple combination rules used at the same time, the data object set specified by each combination rule is different, but there may be part of the same data Object set.
- the foregoing only shows the generation process of a combination instance. In fact, in what order is the data object extracted from each data object set to be combined with the first data object, this embodiment does not specifically limit it, as long as The data object combination that meets the quantity requirement is finally obtained, and these data object combinations are all different.
- step 103 of the method provided in this embodiment "from the multiple data object combinations, select a data object combination that meets the matching rules as an example of clothing matching for the user's reference.”
- the attributes of the data object include at least one of the following attribute items: type, style, color information, texture information, sales records, additional purchase rate, click rate, and shelf time.
- type, style, etc. of the attributes of the data object can be obtained from the information filled in by the merchant; the color information and texture information can be identified by the corresponding algorithm, and the sales record, additional purchase rate, click-through rate and shelf time can be obtained. Obtained from the server.
- This embodiment does not specifically limit the method for obtaining the attribute of the data object, and the corresponding technology in the prior art can be used to obtain it.
- the aforementioned filtering rules may include, but are not limited to: rules for styles, rules for colors, rules for bright spots, and so on.
- step 1032 may include: using style-specific rules to determine whether the attribute combination of the data object contained in the first data object combination among the multiple data object combinations meets all State the rules for styles. Further, the step of "using the style-specific rule to determine whether the attribute combination of the data object contained in the first data object combination among the plurality of data object combinations meets the style-specific rule" may specifically be:
- step 1033 "deleting the data object combination whose attribute combination of the data object contained in the multiple data object combination meets the filtering rule" includes:
- the first data object combination is deleted.
- the above step 1032 may include: using the color-specific rule to determine whether the attribute combination of the data object contained in the first data object combination among the multiple data object combinations meets all State the rules for styles. Further, the step of "using the color-specific rule to determine whether the attribute combination of the data object contained in the first data object combination of the plurality of data object combinations meets the style-specific rule" may specifically include:
- step 1033 "deleting the data object combination whose attribute combination of the data object contained in the multiple data object combination meets the filtering rule" includes:
- the first data object combination is deleted.
- the above step 1032 may include: using the rule for a bright spot, determining whether the attribute combination of the data object contained in the first data object combination among the plurality of data object combinations meets the target Bright spot rules. Further, the step of "using the rule for bright spots to determine whether the attribute combination of the data objects contained in the first data object combination among the plurality of data object combinations meets the rule for bright spots" may specifically include:
- step 1033 "deleting the data object combination whose attribute combination of the data object contained in the multiple data object combination meets the filtering rule" includes:
- the first data object combination is deleted.
- the technical solution provided by this embodiment can be simply summarized as: a process of data object combination construction-filtering-sorting.
- the clothing knowledge of clothing experts or experts can be modeled into the execution logic of these three links, which realizes universal automatic calculation and output of matching.
- the combination rule is introduced in the data object combination structure part, and the potential clothing matching example candidates are quickly constructed in the huge product arrangement and combination space to solve the computational complexity problem.
- the filtering rules are used to accurately filter out the clothing matching examples that meet the decent and aesthetic requirements in daily wear from the combination of multiple data objects, so as to solve the problem of matching quality.
- the data object combination structure part it is based on the combination rules deposited by clothing experts or masters to guide the user to request the product and which types of products are combined to form a data object combination.
- the user requested a dress.
- the dress can be empty or jacket for the top part, and empty or trousers for the bottom part. Shoes and bags are necessary.
- the 4 parts of the commodity (coat, pants, shoes and bag) in the candidate commodity pool can be called out in the storage module, such as 200 pieces of each kind of commodity, as the data object set of each kind of commodity. Generate collocation candidates.
- the filtering part can use filtering rules to filter out inappropriate data object combinations that do not conform to aesthetic principles in color and texture.
- the filtering rules may include, but are not limited to, at least one of the following: filtering rules for styles, filtering rules for colors, filtering rules for bright spots, and so on.
- the filtering rules for styles can be simply understood as the style matching relationship represented in the table shown in FIG. 4.
- FIG. 4 Refer to the table format shown in Figure 4, which is composed of combinations of styles precipitated by experts or experts that cannot be worn at the same time, which can generate 2000 or more grid points; Figure 4 only shows the matching relationship of some styles.
- the role of the filtering rules for styles is to filter out data object combinations that do not meet the appropriate requirements in terms of composition structure and seasonal temperature.
- the data object combination is usually a set of 3 to 5 pieces.
- the data object combination is checked in pairs for all the products contained in the data object combination. If a combination that does not meet the filtering rules for styles is encountered, the data object combination needs to be deleted.
- a certain data object combination includes: coat, top and trousers; correspondingly, query whether the coat and top meet the filtering rules for styles, whether the top and trousers meet the filtering rules for styles, and whether the jackets and trousers meet Filtering rules for styles. If a combination of jacket and top, top and trousers, and jacket and trousers is a combination that is not suitable for wearing at the same time through the query, the data object combination needs to be filtered out, that is, deleted from multiple data object combinations .
- the color information of each product will contain 1 to 3 colors
- the texture information will contain 1 to 3 texture tags, as follows:
- Color information (Color1,AreaC1), (Color2,AreaC2)...
- Texture information (Texture1,AreaT1), (Texture2,AreaT2)...
- Color is a color label, such as carmine
- the corresponding Area is the proportion of the area of the color label in the product, such as 40%
- Texture is a texture label, such as polka dots
- the corresponding Area is the texture label in The area percentage of the product, such as 70%.
- the color filtering rule can be specifically as follows: more than 3 main colors in the data object combination need to be filtered out.
- the implementation of this rule is to traverse each product in a data object combination in turn, and count the number of dominant colors. When the number of dominant colors of different types is greater than 3, the data object combination needs to be filtered out, that is, from multiple Deleted from the combination of data objects.
- the filtering rule for bright spots may be specifically as follows: more than one bright spot in the data object combination needs to be filtered out.
- the implementation of this rule is to traverse each product in a data object combination in turn, and count the number of bright spots. When the total number of bright spots of different types is greater than 1, the data object combination needs to be filtered out, that is, from multiple data objects Delete from the combination.
- the remaining data object combinations that are not filtered out are examples of clothing matching that can be referenced by the user.
- the remaining half can be provided to users as examples of clothing matching.
- these clothing matching examples can be sorted to determine the final display order for users.
- this step is to further optimize the click-through rate of the match.
- the matching score can be calculated according to the popularity score of each product in each clothing matching instance (which can be sales, click-through rate, additional purchase rate, etc., which can be adjusted according to different business goals), and then sorted according to the matching score. Then display it to users in order from high to low.
- FIG. 5 shows a schematic flowchart of a data object processing method provided by another embodiment of the present application. As shown in the figure, the data object processing method includes:
- the attribute of the first data object and the attribute of the at least one second data object determine whether the at least one second data object is matched with the first data object.
- each type of second data object in at least one type of second data object can be obtained in a corresponding type of data object set.
- the first data object and the second data object may be clothing products, such as dresses, coats, shoes, bags, and so on.
- the first data object and the second data object may be a product number, a product name, etc. used to uniquely identify a product.
- the attribute may include at least one of the following: location, style, color information, texture information, sales record, additional purchase rate, click rate, and shelf time.
- this step 202 may include at least one of the following implementation steps:
- the style of the first data object and the style of any second data object in the at least one second data object are determined whether the first data object and the at least one second data object match;
- the technical solution provided in this embodiment extracts a second data object to be combined with the first data object from a set of at least one type of data object that is different from the type of the first data object to generate multiple data object combinations, and then use Matching rules, rather than strict matching templates, select clothing matching examples from multiple data object combinations that can be referenced by users. It is universally applicable and produces more rich combinations under the premise of ensuring matching quality.
- the technical solutions provided by the embodiments of this application can produce multiple sets of matching suggestions for 99% of the products in the target product pool, and the satisfaction rate of manual evaluation reaches 80%.
- the technical solutions provided by the embodiments of the present application can be applied in a clothing matching recommendation scenario.
- a user selects a dress through a client application (APP), and adds the selected dress to a shopping cart.
- the client monitors the user's behavior, it recommends coats, shoes, bags, etc. that can be matched with the dress; and displays the page information corresponding to these coats, shoes, and bags that can be matched with the dress as recommended information on the user interface on.
- the client sends a matching request to the server for the dress.
- the server can use the methods provided in the foregoing embodiments to feed back to the client the coats, shoes, bags, etc. that can be recommended for the user. That is, the technical solution provided by this embodiment can also be implemented by the following hardware system architecture.
- the clothing matching information processing system includes:
- the client 301 is configured to send request information to the server in response to the operation performed by the user on the first data object; receive the clothing matching instance fed back by the server for the first data object; and provide the clothing matching instance To the user;
- the server 302 is configured to obtain request information sent by the client for the first data object; extract from a set of at least one type of data object that is different from the type of the first data object to be used with the first data object.
- the second data object of the object combination to generate a plurality of data object combinations; and the data object combination that meets the matching rules among the plurality of data object combinations is fed back to the client as a clothing matching instance.
- the technical solution provided in this embodiment extracts a second data object to be combined with the first data object from a set of at least one type of data object that is different from the type of the first data object to generate multiple data object combinations, and then use Matching rules, rather than strict matching templates, select clothing matching examples from multiple data object combinations that can be referenced by users. It is universally applicable and produces more rich combinations under the premise of ensuring matching quality.
- the technical solutions provided by the embodiments of this application can produce multiple sets of matching suggestions for 99% of the products in the target product pool, and the satisfaction rate of manual evaluation reaches 80%.
- FIG. 7 shows a schematic flowchart of the method for processing the clothing matching information provided by an embodiment of the present application.
- the execution subject of the method provided in this embodiment may be a client, which may be a mobile phone, a tablet computer, a PDA (Personal Digital Assistant), a POS (Point of Sales, sales terminal), a car computer, etc. Any terminal device.
- the method includes:
- the clothing matching instance is a combination of data objects that meet the matching rules selected from a combination of multiple data objects; the combination of multiple data objects is a combination of at least one type that is different from the type of the first data object. Is generated by combining the second data object extracted from the data object set with the first data object.
- the designated operation may include: the user clicks the corresponding control key, the user makes a control voice for the first data object, or the user makes a designated action, which is not specifically limited in this embodiment.
- the technical solution provided in this embodiment extracts a second data object to be combined with the first data object from a set of at least one type of data object that is different from the type of the first data object to generate multiple data object combinations, and then use Matching rules, rather than strict matching templates, select clothing matching examples from multiple data object combinations that can be referenced by users. It is universally applicable and produces more rich combinations under the premise of ensuring matching quality.
- the technical solutions provided by the embodiments of this application can produce multiple sets of matching suggestions for 99% of the products in the target product pool, and the satisfaction rate of manual evaluation reaches 80%.
- step 403 "provide the clothing matching instance to the user” may include at least one of the following:
- FIG. 8 shows a schematic flowchart of a method for processing clothing matching information provided by an embodiment of the present application.
- the execution subject of the method provided in this embodiment may be a server, and the server may be a common server, a virtual server, or the cloud, etc., which is not specifically limited in this embodiment.
- the method includes:
- the method provided in this embodiment can also implement other steps described in the above embodiments.
- the method provided in this embodiment can also implement other steps described in the above embodiments.
- the technical solution provided in this embodiment extracts a second data object to be combined with the first data object from a set of at least one type of data object that is different from the type of the first data object to generate multiple data object combinations, and then use Matching rules, rather than stringent matching templates, select clothing matching examples from multiple data object combinations that can be used for users to refer to.
- the universality is strong, and the matching output is more rich under the premise of ensuring matching quality
- the technical solutions provided by the embodiments of this application can produce multiple sets of matching suggestions for 99% of the products in the target product pool, and the satisfaction rate of manual evaluation reaches 80%.
- FIG. 9 shows a schematic structural diagram of an apparatus for processing clothing matching information provided by an embodiment of the present application.
- the data object processing device includes: an acquisition module 11, a combination module 12 and a selection module 13.
- the acquisition module 11 is used to acquire at least one type of data object set different from the type of the first data object
- the combination module 12 is used to extract from the at least one type of data object set to be used with the The second data object of the first data object combination is used to generate multiple data object combinations
- the selection module 13 is used to select a data object combination that meets the matching rules from the multiple data object combinations as a reference for the user Examples of clothing matching.
- the technical solution provided in this embodiment extracts a second data object to be combined with the first data object from a set of at least one type of data object that is different from the type of the first data object to generate multiple data object combinations, and then use Matching rules, rather than stringent matching templates, select clothing matching examples from multiple data object combinations that can be used for users to refer to.
- the universality is strong, and the matching output is more rich under the premise of ensuring matching quality
- the technical solutions provided by the embodiments of this application can produce multiple sets of matching suggestions for 99% of the products in the target product pool, and the satisfaction rate of manual evaluation reaches 80%.
- combination module 12 may be specifically used for:
- one data object is selected as a second data object from a part of the data object set or all data object sets in the data object set of the at least one type, and the first data object is combined into a data object Combination; until the number of different data object combinations generated meets the preset number requirements.
- the at least one combination rule includes: a first combination rule and a second combination rule; and the combination module 12 is further configured to:
- selection module 13 is also used for:
- the data object combination whose attribute combination of the data object contained in the multiple data object combination meets the filtering rule is deleted.
- the attributes of the data object include at least one of the following attribute items: type, style, color information, texture information, sales record, additional purchase rate, click rate, and shelf time.
- the filtering rules include: style-specific rules; correspondingly, the selection module 13 is also used for:
- the selection module 13 is further configured to: when the style combination of the first data object and the style combination of any second data object in the at least one second data object meet the style-specific rule, change The first data object combination is deleted.
- filtering rules include: rules for colors; and the selection module 13 is also used for:
- the selection module is further configured to delete the first data object combination when the number of dominant colors is greater than a first threshold.
- filtering rules include: rules for bright spots; and the selection module 13 is also used for:
- selection module 13 is also used for:
- the first data object combination is deleted.
- the acquisition module is also used for:
- a plurality of data objects in the first order are selected to form the first type of data object set.
- the method provided in this embodiment further includes: a sorting module.
- the sorting module is used for sorting a plurality of clothing matching instances; according to the sorting results, providing a plurality of clothing matching instances to the user.
- the apparatus for processing clothing matching information provided in the foregoing embodiment can implement the technical solutions described in the foregoing method embodiments.
- the foregoing modules or units please refer to the corresponding content in the foregoing method embodiments. , I won’t repeat it here.
- FIG. 10 shows a schematic structural diagram of a data object processing apparatus provided by an embodiment of the present application.
- the data object processing device includes: an acquisition module 21, a determination module 22, and a provision module 23.
- the obtaining module 21 is used to obtain at least one type of second data object that is different from the type of the first data object;
- the determining module 22 is used to obtain according to the attributes of the first data object and the at least one first data object. 2.
- the attribute of the data object determining whether the at least one second data object is matched with the first data object; the providing module 23 is used when the at least one second data object is matched with the first data object, The at least one second data object is provided to the user.
- the technical solution provided in this embodiment extracts a second data object to be combined with the first data object from a set of at least one type of data object that is different from the type of the first data object to generate multiple data object combinations, and then use Matching rules, rather than strict matching templates, select clothing matching examples from multiple data object combinations that can be referenced by users. It is universally applicable and produces more rich combinations under the premise of ensuring matching quality.
- the technical solutions provided by the embodiments of this application can produce multiple sets of matching suggestions for 99% of the products in the target product pool, and the satisfaction rate of manual evaluation reaches 80%.
- the first data object and the second data object are specific clothing products, and correspondingly, the attributes of the first data object and the second data object may specifically include at least one of the following: , Style, color information, texture information, sales records, additional purchase rate, click rate, and shelf time; correspondingly, the determination module 22 is also used for:
- the number of the main colors determine whether the first data object and the at least one second data object match; and/or
- the color information of the first data object and the color information of the at least one second data object determine whether there is a color bright spot as the first determination result; according to the texture information of the first data object and the at least one second data object The texture information of the data object determines whether there is a texture bright spot as the second determination result, and according to the first determination result and the second determination result, it is determined whether the first data object and the at least one second data object match.
- FIG. 11 shows a schematic structural diagram of an apparatus for processing clothing matching information provided by an embodiment of the present application.
- the data object processing device includes: a sending module 31, a receiving module 32, and an output module 33.
- the sending module 31 is configured to send request information to the server in response to the operation performed by the user on the first data object;
- the receiving module 32 is configured to receive the clothing matching instance that the server feedbacks on the first data object;
- the output module 33 is used to provide the clothing matching examples to the user.
- the clothing matching instance is a combination of data objects that meet the matching rules selected from a combination of multiple data objects;
- the combination of multiple data objects is a combination of at least one type that is different from the type of the first data object. Is generated by combining the second data object extracted from the data object set with the first data object.
- the output module is also used for:
- the technical solution provided in this embodiment extracts a second data object to be combined with the first data object from a set of at least one type of data object that is different from the type of the first data object to generate multiple data object combinations, and then use Matching rules, rather than strict matching templates, select clothing matching examples from multiple data object combinations that can be referenced by users. It is universally applicable and produces more rich combinations under the premise of ensuring matching quality.
- the technical solutions provided by the embodiments of this application can produce multiple sets of matching suggestions for 99% of the products in the target product pool, and the satisfaction rate of manual evaluation reaches 80%.
- the apparatus for processing clothing matching information provided in the foregoing embodiment can implement the technical solutions described in the foregoing method embodiments.
- the foregoing modules or units please refer to the corresponding content in the foregoing method embodiments. , I won’t repeat it here.
- FIG. 12 shows a schematic structural diagram of an apparatus for processing clothing matching information provided by an embodiment of the present application.
- the data object processing device includes: an acquisition module 41, an extraction module 42, and a feedback module 43.
- the acquiring module 41 is used to acquire the request information sent by the client for the first data object;
- the extracting module 42 is used to collect data objects of at least one type different from the type of the first data object. , Extracting a second data object to be combined with the first data object to generate a plurality of data object combinations;
- the feedback module 43 is configured to use the data object combination that meets the collocation rule among the plurality of data object combinations as The clothing matching examples are fed back to the client.
- the technical solution provided in this embodiment extracts a second data object to be combined with the first data object from a set of at least one type of data object that is different from the type of the first data object to generate multiple data object combinations, and then use Matching rules, rather than strict matching templates, select clothing matching examples from multiple data object combinations that can be referenced by users. It is universally applicable and produces more rich combinations under the premise of ensuring matching quality.
- the technical solutions provided by the embodiments of this application can produce multiple sets of matching suggestions for 99% of the products in the target product pool, and the satisfaction rate of manual evaluation reaches 80%.
- the apparatus for processing clothing matching information provided in the foregoing embodiment can implement the technical solutions described in the foregoing method embodiments.
- the foregoing modules or units please refer to the corresponding content in the foregoing method embodiments. , I won’t repeat it here.
- FIG. 13 shows a schematic structural diagram of an electronic device provided by an embodiment of the present application.
- the electronic device includes a memory 51 and a processor 52.
- the memory 51 may be configured to store other various data objects to support operations on the electronic device. Examples of these data objects include instructions for any application or method operating on an electronic device.
- the memory 51 can be implemented by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable and Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic Disk or Optical Disk.
- SRAM static random access memory
- EEPROM electrically erasable programmable read-only memory
- EPROM erasable and Programmable Read Only Memory
- PROM Programmable Read Only Memory
- ROM Read Only Memory
- Magnetic Memory Flash Memory
- Magnetic Disk Magnetic Disk or Optical Disk.
- the processor 52 is coupled with the memory 51, and is configured to execute the program stored in the memory 51 for:
- a data object combination that meets the matching rules is selected as a clothing matching example that can be referenced by the user.
- processor 52 executes the program in the memory 51, in addition to the above functions, it may also implement other functions. For details, please refer to the descriptions of the previous embodiments.
- the electronic device further includes: a display 54, a communication component 53, a power supply component 55, an audio component 56 and other components. Only part of the components are schematically shown in FIG. 13, which does not mean that the electronic device only includes the components shown in FIG.
- An embodiment of the present application also provides an electronic device.
- the structure of the electronic device provided in this embodiment is similar to the structure of the foregoing electronic device embodiment, as shown in FIG. 13.
- the electronic device includes a memory and a processor.
- the processor is coupled with the memory, and is configured to execute the program stored in the memory for:
- the at least one second data object is matched with the first data object, the at least one second data object is provided to the user.
- An embodiment of the present application also provides a client device.
- the structure of the client device provided in this embodiment is similar to the structure of the foregoing electronic device embodiment, as shown in FIG. 13.
- the client device includes a memory and a processor.
- the processor is coupled with the memory, and is configured to execute the program stored in the memory for:
- the clothing matching instance is a combination of data objects that meet the matching rules selected from a combination of multiple data objects; the combination of multiple data objects is a combination of at least one type that is different from the type of the first data object. Is generated by combining the second data object extracted from the data object set with the first data object.
- An embodiment of the present application also provides a server device.
- the structure of the server device provided in this embodiment is similar to the structure of the foregoing electronic device embodiment, as shown in FIG. 13.
- the server device includes a memory and a processor.
- the processor is coupled with the memory, and is configured to execute the program stored in the memory for:
- the data object combination that meets the matching rule among the multiple data object combinations is fed back to the client as a clothing matching instance.
- an embodiment of the present application also provides a computer-readable storage medium storing a computer program, which can realize the steps or functions of the clothing matching information processing method provided by the foregoing embodiments when the computer program is executed by a computer.
- the embodiments of the present application also provide a computer-readable storage medium storing a computer program, which can realize the steps or functions of the data object processing method provided by the foregoing embodiments when the computer program is executed by a computer.
- the device embodiments described above are merely illustrative.
- the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in One place, or it can be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments. Those of ordinary skill in the art can understand and implement without creative work.
- each implementation manner can be implemented by software plus a necessary general hardware platform, and of course, it can also be implemented by hardware.
- the computer software product can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic A disc, an optical disc, etc., include a number of instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiment.
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
一种服饰搭配信息的处理方法、数据对象处理方法、系统及设备。其中,服饰搭配信息的处理方法包括如下的步骤:获取与第一数据对象所属类型不同的至少一个类型的数据对象集(101);从所述至少一个类型的数据对象集中提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合(102);从所述多个数据对象组合中,选出符合搭配规则的数据对象组合作为能供用户参考的服饰搭配实例(103)。该方法普适性强,在保证搭配质量的前提下产出的搭配更具丰富性,且具有较好的搭配满意度。
Description
本申请要求2019年01月31日递交的申请号为201910101037.6、发明名称为“服饰搭配信息的处理方法、数据对象处理方法、系统及设备”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
本发明涉及计算机技术领域,尤其涉及一种服饰搭配信息的处理方法、数据对象处理方法、系统及设备。
长久以来,人们一直习惯在琳琅满足、款式与色彩众多的鞋服包商品中通过反复挑选来选择适合自己的商品。大多数的普通用户不太知道如何搭配这些商品,如连衣裙搭配什么样的鞋子,搭配什么款式的包等。
目前,电商平台推荐商品的常用逻辑是基于用户行为的推荐技术,给用户推荐的结果大多是与已购买、收藏或浏览的商品同属一个类目的商品。比如,用户对连衣裙感兴趣,在平台经常浏览和点击连衣裙,在推荐流量窗口就会展示多种类似的连衣裙。这种推荐方式会带来重复推荐的结果,用户明明已购买了商品,系统却还重复推荐同类商品,容易造成用户的困扰,用户体验不好。
发明内容
本申请各实施例提供一种解决或部分地解决上述问题的服饰搭配信息的处理方法、数据对象处理方法、系统及设备。
在本申请的一个实施例中,提供了一种服饰搭配信息处理方法。该方法包括:
获取与第一数据对象所属类型不同的至少一个类型的数据对象集;
从所述至少一个类型的数据对象集中提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;
从所述多个数据对象组合中,选出符合搭配规则的数据对象组合作为能供用户参考的服饰搭配实例。
在本申请的另一个实施例中,提供了一种数据对象处理方法。该方法包括:
与第一数据对象相关的指定事件发生时,获取与所述第一数据对象所属类型不同的 至少一个类型的第二数据对象;
根据所述第一数据对象的属性及所述至少一个第二数据对象的属性,判定所述至少一个第二数据对象是否与所述第一数据对象搭配;
所述至少一个第二数据对象与所述第一数据对象搭配时,将所述至少一个第二数据对象提供给用户。
在本申请的又一个实施例中,提供了一种服饰搭配信息的处理方法。该方法包括:
响应于用户针对第一数据对象进行的操作,向服务端发送请求信息;
接收所述服务端针对所述第一数据对象反馈的服饰搭配实例;
将所述服饰搭配实例提供给所述用户;
其中,所述服饰搭配实例是从多个数据对象组合中选出的符合搭配规则的数据对象组合;所述多个数据对象组合是将从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中提取的第二数据对象与所述第一数据对象组合生成的。
在本申请的又一个实施例中,提供了一种服饰搭配信息的处理方法。该方法包括:
获取客户端针对所述第一数据对象发送的请求信息;
从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中,提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;
将所述多个数据对象组合中符合搭配规则的数据对象组合作为服饰搭配实例反馈至所述客户端。
在本申请的一个实施例中,提供了一种服饰搭配信息的处理系统。该系统包括:
客户端,用于响应于用户针对第一数据对象进行的指定操作,向服务端发送所述指定操作对应的请求信息;接收所述服务端针对所述第一数据对象反馈的服饰搭配实例;将所述服饰搭配实例提供给所述用户;
服务端,用于获取客户端针对所述第一数据对象发送的请求信息;从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中,提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;将所述多个数据对象组合中符合搭配规则的数据对象组合作为服饰搭配实例反馈至所述客户端。
在本申请的一个实施例中,提供了一种电子设备。该电子设备包括存储器及处理器;其中,
所述存储器,用于存储程序;
所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:
获取与第一数据对象所属类型不同的至少一个类型的数据对象集;
从所述至少一个类型的数据对象集中提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;
从所述多个数据对象组合中,选出符合搭配规则的数据对象组合作为能供用户参考的服饰搭配实例。
在本申请的另一个实施例中,提供了一种电子设备。该电子设备包括存储器及处理器;其中,
所述存储器,用于存储程序;
所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:
获取与第一数据对象所属类型不同的至少一个类型的第二数据对象;
根据所述第一数据对象的属性及所述至少一个第二数据对象的属性,判定所述至少一个第二数据对象是否与所述第一数据对象搭配;
所述至少一个第二数据对象与所述第一数据对象搭配时,将所述至少一个第二数据对象提供给用户。
在本申请的一个实施例中,提供了一种客户端设备。所述客户端设备包括存储器及处理器;其中,
所述存储器,用于存储程序;
所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:
响应于用户针对第一数据对象进行的操作,向服务端发送请求信息;
接收所述服务端针对所述第一数据对象反馈的服饰搭配实例;
将所述服饰搭配实例提供给所述用户;
其中,所述服饰搭配实例是从多个数据对象组合中选出的符合搭配规则的数据对象组合;所述多个数据对象组合是将从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中提取的第二数据对象与所述第一数据对象组合生成的。
在本申请的一个实施例中,提供了一种服务端设备。所述服务端设备包括存储器及处理器;其中,
所述存储器,用于存储程序;
所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:
获取客户端针对所述第一数据对象发送的请求信息;
从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中,提取用以与 所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;
将所述多个数据对象组合中符合搭配规则的数据对象组合作为服饰搭配实例反馈至所述客户端。
本申请实施例提供的技术方案,从与第一数据对象所属类型不同的至少一个类型的数据对象集中提取用以与第一数据对象组合的第二数据对象,以生成多个数据对象组合,然后利用搭配规则,而非条件严苛的搭配模板,从多个数据对象组合中选出能供用户参考的服饰搭配实例,普适性强,在保证搭配质量的前提下产出的搭配更具丰富性;另外,通过效果实测,本申请实施例提供的技术方案能够为目标商品池99%的商品产出多套的搭配建议,并且人工评测满意率达到80%。
为了更清楚地说明本申请实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作一简单地介绍,显而易见地,下面描述中的附图是本申请的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1为本申请一实施例提供的服饰搭配信息的处理方法的流程示意图;
图2为本申请一实施例以连衣裙为例对服饰搭配信息的处理方法进行原理性描述的数据对象流转示图;
图3为本申请一实施例提供的组合规则示意图;
图4为本申请一实施例为表征款式搭配关系的表格;
图5为本申请另一实施例提供的数据对象处理方法的流程示意图;
图6为本申请一实施例提供的服饰搭配信息的处理系统的结构示意图;
图7为本申请又一实施例提供的服饰搭配信息的处理方法的流程示意图;
图8为本申请又一实施例提供的服饰搭配信息的处理方法的流程示意图;
图9为本申请一实施例提供的服饰搭配信息的处理装置的结构示意图;
图10为本申请另一实施例提供的数据对象处理装置的结构示意图;
图11为本申请又一实施例提供的服饰搭配信息的处理装置的结构示意图;
图12为本申请又一实施例提供的服饰搭配信息的处理装置的结构示意图;
图13为本申请一实施例提供的电子设备的结构示意图。
为了使本技术领域的人员更好地理解本申请方案,下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚、完整地描述。
在本申请的说明书、权利要求书及上述附图中描述的一些流程中,包含了按照特定顺序出现的多个操作,这些操作可以不按照其在本文中出现的顺序来执行或并行执行。操作的序号如101、102等,仅仅是用于区分各个不同的操作,序号本身不代表任何的执行顺序。另外,这些流程可以包括更多或更少的操作,并且这些操作可以按顺序执行或并行执行。需要说明的是,本文中的“第一”、“第二”等描述,是用于区分不同的消息、设备、模块等,不代表先后顺序,也不限定“第一”和“第二”是不同的类型。另外,如下各实施例仅仅是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
现有技术通常采用如下三种方案实现服饰搭配推荐。下面将分别进行说明:
第一种、基于手工编辑
基于手工编辑方案,质量高,但搭配数量极少。通过发动专家和达人等群体手工制作搭配,质量很高,但受制于人力效率,通常日产量在百千量级,数量极少。平台商品在亿万量级,因此大量的商品是得不到人工的搭配建议的。与用户本身已购商品无关的搭配,无法对用户发挥作用。
第二种、基于关联购买
基于关联购买的方案,质量一般,数量较多。通过挖掘用户在一段时间窗口内的共同购买的商品,形成搭配样本,训练模型来自动产出搭配。该方法会遇到训练样本噪声太大的问题,因为通常用户在一段时间内购买的服装商品并不保证为同一身穿搭所采购,因此训练样本质量不高(或将样本加工到符合搭配质量要求的成本过高),导致模型效果有限;该方法仅适合做关联商品推荐。
第三种、基于模板
基于搭配模板的方案,质量较高,在开放商品集下数量多,在限定商品集下数量较少。针对第一种方案的商品极少的问题,将手工搭配作为模板,利用单品找相似功能,将一套搭配扩展成多套相似搭配,该方法能在保证一定搭配质量的前提下,有效提升搭配数量。然而,平台能积累的(手工)搭配模板通常在百万以下的量级,相比动辄上千万甚至过亿的平台商品(能排列组合成的搭配数量是几近无穷的),所能满足的商品以 及能覆盖的搭配形式仍是有限的。一个搭配模板对商品的丰富性要求是很高的。比如一个4件套搭配模板,意味着要求特定的上装、下装、鞋和包4个部位的商品都符合相似的条件下,才构成该模板下合格的搭配。因此,特别是求在限定商品集下产出搭配(比如某个品牌、某个店铺内)时,由于候选商品有限,不足以满足特定模板所有部位都相似的要求,导致搭配产出失败。综上,模板式方法能生产的搭配仍然有限,无法覆盖平台所有商品的搭配需求。
为给用户提供更丰富、满意度好的搭配结果,需克服的技术挑战有:
A.计算复杂性
一套搭配通常由3~5个部位商品组成。对于电商平台来说,服装、鞋和包通常涵盖40个以上类型,每个类型的商品候选多达上千万。因此一套搭配的商品候选搜索空间是巨大的,可以作为组合候选的商品排列组合不可穷尽。
B.商品覆盖率
在有充足候选商品集的前提下,如果方法无法为某个商品产出搭配,将错失一次用户展现搭配和商品转化的机会。因此,对于平台海量的服装商品,方法的覆盖率应该达到95%以上。
C.搭配质量
如果围绕用户请求商品展示的搭配建议不符合基本的服装穿搭审美需求,用户同样不会继续浏览和点击。因此,要求方法在保证产出搭配的覆盖率和丰富性的同时,要求较高的合格率。
D.实现效率
对于电商平台海量服装商品的中的每一件均提供搭配服务,必须能提供实时计算能力,而不是提前缓存计算结果(枚举数量太大成本过高)。因此,对于单个商品返回10套及以上搭配的计算时间要求小于200ms。
图1示出了本申请一实施例提供的服饰搭配信息的处理方法的流程示意图。本实施例提供方案的执行主体可以数据对象处理装置,该装置可以是集成在客户端或服务端上的一个具有嵌入式程序的硬件,也可以是安装在客户端或服务端中的一个应用软件,还可以是嵌入在客户端或服务端操作系统中的工具软件等,本申请实施例对此不作限定。其中,客户端可以为包括手机、平板电脑、PDA(Personal Digital Assistant,个人数字助理)、POS(Point of Sales,销售终端)、车载电脑等任意终端设备。服务端可以服务器、云端等;本实施例对此不作具体限定。如图1所示,所述方法包括:
101、获取与第一数据对象所属类型不同的至少一个类型的数据对象集。
102、从所述至少一个类型的数据对象集中提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合。
103、从所述多个数据对象组合中,选出符合搭配规则的数据对象组合作为能供用户参考的服饰搭配实例。
在服饰类电商场景下,上述第一数据对象、第二数据对象即具体的服饰类商品,如连衣裙、外套、鞋、包等等。第一数据对象、第二数据对象可以是用于唯一标识商品的商品号或商品名称等等。对于客户端来说,上述各步骤可在用户进行了如下任意一种行为后被触发执行:
用户点击了请求查看商品A的服饰搭配实例的控件;
用户发出请求为商品A推荐服饰搭配实例的语音;
用户将商品A加入购物车;
用户关注商品A;
用户购买了商品A;等等。
上述用户行为的对象—商品A,即本实施例中的第一数据对象。
对于服务端来说,服务端可在接收到客户端针对第一数据对象发送的搭配请求后,执行上述各步骤。
本实施例可应用于很多应用场景,比如服饰类类电商场景。第一数据对象所属类型即该商品的类型,如鞋类、服装类、包类、饰品(如项链、头饰、耳环、戒指等)类等等。服装类又因为穿着部位的不同,分为:上装类、下装类、连体类等等。连体类又可分为:连衣裙类、连体裤类等等。同样的,饰品类因为佩戴部位的不同,可分为:项链类、头饰类、耳环类、戒指类等等。
假设在服饰类电商场景,第一数据对象为服装信息,该服装信息所属类型为连衣裙,相应的,可获取如下至少一种类型的数据对象集:
鞋类数据对象集、饰品类数据对象集、上装类数据对象集、包类数据对象集、下装类数据对象集等等。
例如,获取鞋类的数据对象集及包类的数据对象集。具体实施时,各类数据对象集中包含的数据对象(即商品)可以是热门或新上架的商品。在一种可实现的技术方案中,本实施例提供的技术方案,还可包括如下步骤:
收集同一类型的数据对象;
按照排序规则,对收集到的数据对象进行排序;
选取排在前的多个数据对象,得到该类型的数据对象集。
例如,排序规则可具体为:按人气分+商品上架时间倒排。选取的数据对象的数量可人为设定,比如,排在前100、150、200或更多,本实施例对此不作具体限定。
数据对象所属类型(如鞋类、服装类等)、排序参数(如人气分、上架时间等等)均可记载在数据对象的属性中。即,以与第一数据对象所属类型不同的第一类型的数据对象集为例,对获取第一类型的数据对象集的过程进行说明,具体的包括:
获取所述第一类型数据对象池中各数据对象的属性;
根据所述第一类型数据对象池中各数据对象的属性,对所述第一类型数据对象池中的数据对象进行排序;
选取排序在前的多个数据对象组成所述第一类型的数据对象集。
其中,所述数据对象的属性可包括:类型、款式、颜色信息、纹理信息、销售记录、加购率、点击率、上架时间、适合季节、适合场合、适合风格等。人气分可通过销售记录、加购率及点击率等计算得到。
上述102中,当有一个类型的数据对象集时,可直接从数据对象集中提取第二数据对象,与第一数据对象组合得到多个数据对象组合。例如,第一数据对象为连衣裙,步骤101获取到一个鞋类的数据对象集,该鞋类的数据对象集中含有200个数据对象,可将数据对象集中包含的200个数据对象分别与连衣裙组合,得到200个数据对象组合。
当有两个或两个以上的数据对象集时,可从部分或全部数据对象集中提取第二数据对象,与第一数据对象组合得到多个数据对象组合。例如,第一数据对象为连衣裙,步骤101获取到鞋类的数据对象集、包类数据对象集及外套类数据对象集。具体实施时,可在鞋类数据对象集中提取一个作为第二数据对象的数据对象a,从包类数据对象集中提取一个作为第二数据对象的数据对象b,然后将数据对象a、数据对象b与第一数据对象组合成一个数据对象组合;或者,可在鞋类数据对象集中提取一个作为第二数据对象的数据对象c,从包类数据对象集中提取一个作为第二数据对象的数据对象d,从外套类数据对象集中提取一个作为第二数据对象的数据对象e,然后将数据对象c、数据对象d及数据对象e与第一数据对象组合成一个数据对象组合;等等。
这里需要说明的是:如何从至少一个类型的数据对象集中提取数据对象生成数据对象组合的方式,本实施例不作具体限定,只要保证数据对象组合包含的所有数据对象中各数据对象的类型不同,且多个数据对象组合均不同。
上述103中,搭配规则可包括一个或多个规则,本实施例对此不作具体限定。在不同应用场景下,搭配规则的设定也会不同。比如,在鞋服包电商场景下,搭配规则可包括:针对款式的规则、针对颜色的规则、针对亮点的规则等等。上述搭配规则可基于专家知识得到。具体的,针对款式的规则用于过滤掉组成结构和季节温度等不搭配的数据对象组合,比如:“T恤”+“羽绒裤”的数据对象组合。针对颜色的规则和针对亮点的规则可用于过滤掉不符合审美要求的数据对象组合;其中审美要求可以专家或达人们默认的穿衣规则,如全身不超过3个主色、不超过1个亮点等等。
另外,假设得到的能供用户参考的服饰搭配实例为多个的情况下,本实施例提供的所述方法还包括:
对多个服饰搭配实例进行排序;
按照排序结果,将多个服饰搭配实例提供给用户。
通过增加上述步骤是为了进一步优化搭配的点击率。具体实施时,可以获取各服饰搭配实例中所含商品的人气分(可以是销量、点击率、加购率等等,依据业务目标不同可以调整),计算各服饰搭配实例的分值。例如,将所含商品中除第一数据对象以外的其他数据对象的人气分的平均值作为最后计算得到的分值。然后,按照分值的大小,将分值大的排在前,分值小排在后的方式,对多个服饰搭配实例进行排序。最后,可按分值从高到低排序展现给用户。
本实施例提供的技术方案,从与第一数据对象所属类型不同的至少一个类型的数据对象集中提取用以与第一数据对象组合的第二数据对象,以生成多个数据对象组合,然后利用搭配规则,而非条件严苛的搭配模板,从多个数据对象组合中选出能供用户参考的服饰搭配实例,普适性强,在保证搭配质量的前提下产出的搭配更具丰富性;既能满足用户的实际搭配需求,又有助于提高电商平台的商品转化率;另外,通过效果实测,本申请实施例提供的技术方案能够为目标商品池99%的商品产出多套的搭配建议,并且人工评测满意率达到80%。
在一种可实现的技术方案中,本实施例提供的所述方法中步骤102:“从所述至少一个类型的数据对象集中提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合”,可采用如下步骤实现:
1021、获取至少一个组合规则;
1022、使用所述至少一个组合规则,从所述至少一个类型的数据对象集中的部分数据对象集或全部数据对象集中分别选出一个数据对象作为第二数据对象与所述第一数据 对象组合成数据对象组合;直至生成的不同数据对象组合的数量满足预置数量要求。
上述1021中,至少一个组合规则可基于第一数据对象所属类型获取得到。如图2和图3所示,假设第一数据对象所属类型为连衣裙类,那么获取到的至少一个组合规则可包括:与鞋和包组合的组合规则1、与外套、鞋和包组合的组合规则2、与裤子、鞋和包组合的组合规则3、与外套、裤子、鞋和包组合的组合规则4等等。其中,图3中符号
表征可不选择。参见图3所示,假设第一数据对象所属类型为上衣类,那么获取到的至少一个组合规则可包括:与裤子、鞋和包组合的组合规则1’、与半身裙、鞋和包组合的组合规则2’、与外套、半身裙、鞋和包组合的组合规则3’等等。
实际应用中,可预先为每种类型对应的组合规则进行设置。这样,用户在针对某一类型的数据对象进行购买、查看或申请搭配请求时,可直接调取出该数据对象所属类型对应的组合规则。
下面以所述至少一个组合规则包括:第一组合规则和第二组合规则为例,对上述步骤1022进行说明。一般情况下,两种类型的数据对象集或两种类型以上的数据对象集才需两种或两种以上的组合规则。即,当步骤101中获取到至少两个类型的数据对象集时,上述步骤1022可具体为:
1022’、使用所述第一组合规则和所述第二组合规则,从至少两个类型的数据对象集中的部分数据对象集或全部数据对象集中分别选出一个数据对象作为第二数据对象与所述第一数据对象组合成数据对象组合;直至生成的不同数据对象组合的数量满足预置数量要求。
具体的,该步骤1022’可采用如下步骤实现:
S1、获取所述至少两个类型的数据对象集中各类型的数据对象集对应的当前排序序号Ni,其中,不同类型的数据对象集的i值不同;
S2、从所述第一组合规则指定所有数据对象集中,分别获取排序为Ni的数据对象作为第二数据对象与所述第一数据对象组合成数据对象组合,并将所述第一组合规则指定的所有数据对象集各自对应的Ni更新为Ni+1;
S3、从所述第二组合规则指定的所有数据对象集中,分别获取排序为Ni的数据对象作为第二数据对象与所述第一数据对象组合成数据对象组合,并将所述第二组合规则指定的所有数据对象集各自对应的Ni更新为Ni+1;
S4、重复上述步骤,直至生成的不同组合实例的数量满足预置数量要求。
上述步骤S1至S4可简单理解为如下过程:
参见图2所示的实例,鞋类的数据对象集3对应的N3=1,包类的数据对象集4对应的N4=1,外套类的数据对象集1对应的N1=1;
首先,获取各数据对象集的Ni;数据对象集3对应的N3=1,数据对象集4对应的N4=1,数据对象集1对应的N1=1。
然后,从第一规则指定的数据对象集3中,获取排序为N3=1的商品;从第一规则指定的数据对象集4中,获取排序为N4=1的商品,组成数据对象组合(31,41);并将N3更新为1+1=2;N4更新为1+1=2;
随后,从第二规则指定的数据对象集3中,获取排序为N3=2的商品,从第二规则指定的数据对象集4中,获取排序为N4=2的商品;从第二规则指定的数据对象集1中,获取排序为N1=1的商品,组成数据对象组合(11、32、42);并将N3、N4和N1均在现有基础上再+1,即N3更新为3、N4更新为3、N1更新为2。
此处仅描述了两种组合规则时,生成不同组合实例的过程。当有第三组合规则、第四组合规则、……等更多组合规则时,原理类型。下面以图2所示的包括4个组合规则的实例。生成多个组合实例的过程可具体为:
使用组合规则1,从数据对象集3提取排序序号N=1的商品31、从数据对象集4中提取排序序号N=1的商品41,将用户购买、浏览或请求搭配的连衣裙00、商品31和商品41组成一个3件套数据对象组合[00、31、41];
使用组合规则2,从数据对象集3提取排序序号N+1=2的商品32、从数据对象集4中提取排序序号N+1=2的商品42、从数据对象集1中提取排序序号N=1的商品11,将连衣裙00、商品32、商品42及商品11组成一个4件套数据对象组合[00、32、42、11];
使用组合规则3,从数据对象集3提取排序序号N+2=3的商品33、从数据对象集4中提取排序序号N+2=3的商品43、从数据对象集2中提取排序序号N=1的商品21,将连衣裙00、商品33、商品43及商品21组成一个4件套数据对象组合[00、33、43、21];
使用组合规则4,从数据对象集3提取排序序号N+3=4的商品34、从数据对象集4中提取排序序号N+3=4的商品44、从数据对象集1中提取排序序号N+1=2的商品12、从数据对象集2中提取排序序号N+1=2的商品22,将连衣裙00、商品34、商品44、商品12、商品22组成一个4件套数据对象组合[00、34、44、12、22]。
将所述排序序号N更新为N+1,并重复上述步骤,直至生成的不同组合实例的数量满足预置数量要求。
具体实施时,预置数量可为100、200、….或更多,本实施例对此不作具体限定。
上述实例中,组合规则1指定的数据对象集包括:数据对象集3和数据对象集4;组合规则2指定的数据对象集包括:数据对象集1、数据对象集3和数据对象集4;组合规则3指定的数据对象集包括:数据对象集2、数据对象集3和数据对象集4;组合规则4指定的数据对象集包括:数据对象集1、数据对象集2、数据对象集3和数据对象集4。这里需要说明的是:在具体实施时,组合规则指定的数据对象集可以为一个或多个;同时使用的多个组合规则中各组合规则指定的数据对象集不同,但可有部分相同的数据对象集。另外,上述仅示出了一种组合实例的生成过程,实际上按照什么样的顺序从各数据对象集中提取数据对象以与第一数据对象组合的方式,本实施例并不做具体限定,只要最后得到的符合数量要求的数据对象组合,且这些数据对象组合均不同即可。
进一步的,本实施例提供的所述方法中步骤103:“从所述多个数据对象组合中,选出符合搭配规则的数据对象组合作为能供用户参考的服饰搭配实例”,可具体采用如下方法实现:
1031、获取所述多个数据对象组合中各数据对象组合所含数据对象的属性。
1032、使用针对所述属性中至少一个属性项的过滤规则,判定所述多个数据对象组合中各数据对象组合所含数据对象的属性组合是否符合所述过滤规则。
1033、将所述多个数据对象组合中所含数据对象的属性组合符合所述过滤规则的数据对象组合删除。
在一具体应用场景中,所述数据对象的属性包括如下至少一种属性项:类型、款式、颜色信息、纹理信息、销售记录、加购率、点击率、上架时间。这里需要说明的是:数据对象的属性中类型、款式等可来自于商家填写的信息得到;颜色信息及纹理信息可通过相应算法识别得到,销售记录、加购率、点击率及上架时间均可从服务器中获取到。本实施例对于数据对象的属性获取方法不作具体限定,可采用现有技术中的相应技术获得。
上述过滤规则可包括但不限于:针对款式的规则、针对颜色的规则、针对亮点的规则等等。
当所述过滤规则包括:针对款式的规则时,上述步骤1032可包括:使用针对款式的规则,判定所述多个数据对象组合中的第一数据对象组合所含数据对象的属性组合是否符合所述针对款式的规则。进一步的,该步骤“使用针对款式的规则,判定所述多个数据对象组合中的第一数据对象组合所含数据对象的属性组合是否符合所述针对款式的规则”可具体为:
获取所述第一数据对象组合所含的所述第一数据对象的款式以及与所述第一数据对象组合的至少一个第二数据对象的款式;
判定所述第一数据对象的款式分别与所述至少一个第二数据对象中各第二数据对象的款式组合是否符合所述针对款式的规则。
相应的,上述步骤1033、“将所述多个数据对象组合中所含数据对象的属性组合符合所述过滤规则的数据对象组合删除”,包括:
所述第一数据对象的款式与所述至少一个第二数据对象中任一第二数据对象的款式组合符合所述针对款式的规则时,将所述第一数据对象组合删除。
当所述过滤规则包括:针对颜色的规则时,上述步骤1032可包括:使用针对颜色的规则,判定所述多个数据对象组合中的第一数据对象组合所含数据对象的属性组合是否符合所述针对款式的规则。进一步的,步骤“使用针对颜色的规则,判定所述多个数据对象组合中的第一数据对象组合所含数据对象的属性组合是否符合所述针对款式的规则”,可具体包括:
获取所述第一数据对象组合所含的所述第一数据对象的颜色信息以及与所述第一数据对象组合的至少一个第二数据对象的颜色信息;
基于所述第一数据对象的颜色信息及所述至少一个第二数据对象的颜色信息,确定满足主色条件的主色数量;
判定所述主色数量是否符合所述针对颜色的规则。
相应的,上述步骤1033、“将所述多个数据对象组合中所含数据对象的属性组合符合所述过滤规则的数据对象组合删除”,包括:
所述主色数量大于第一阈值时,将所述第一数据对象组合删除。
当过滤规则包括:针对亮点的规则时,上述1032步骤可包括:使用针对亮点的规则,判定所述多个数据对象组合中的第一数据对象组合所含数据对象的属性组合是否符合所述针对亮点的规则。进一步的,该步骤“使用针对亮点的规则,判定所述多个数据对象组合中的第一数据对象组合所含数据对象的属性组合是否符合所述针对亮点的规则”可具体包括:
获取所述第一数据对象组合所含的所述第一数据对象的颜色信息和纹理信息以及与所述第一数据对象组合的至少一个第二数据对象的颜色信息和纹理信息;
对所述第一数据对象的颜色信息和所述至少一个第二数据对象的颜色信息进行亮点分析,得到与亮点有关的第一分析结果;
对所述第一数据对象的纹理信息和所述至少一个搭配数据对象的纹理信息进行亮点分析,得到与亮点有关的第二分析结果;
判定所述第一分析结果和所述第二分析结果,是否符合所述针对亮点的规则。
相应的,上述步骤1033、“将所述多个数据对象组合中所含数据对象的属性组合符合所述过滤规则的数据对象组合删除”,包括:
根据所述第一分析结果和所述第二分析结果得出亮点数量大于第二阈值时,将所述第一数据对象组合删除。
基于上述内容,本实施例提供的技术方案可简单总结为:数据对象组合构造—过滤—排序的过程。比如,在鞋服包电商场景中,可将服饰专家或达人的穿衣知识建模成这3个环节中的执行逻辑,实现了普适的搭配自动计算和产出。其中,在数据对象组合构造部分引入组合规则,在庞大的商品排列组合空间中迅速构建潜在的服饰搭配实例候选,解决计算复杂性问题。在过滤部分,采用过滤规则在多个数据对象组合中准确地过滤出符合日常穿搭中得体和审美要求的服饰搭配实例,解决搭配质量的问题。最后,还可使用在线召回+过滤排序的搜索框架,支持快速运算并返回结果,满足实时性的要求;经过实验测试,在数量超过200且类目均衡的商品集上,能产出搭配的商品覆盖率超过99%,搭配满意率超过80%。
数据对象组合构造部分
在数据对象组合构造部分,依据的是服饰专家或达人沉淀的组合规则,来指导用户所请求的商品与哪些类型的商品组合形成数据对象组合。如图2所示的实例,用户请求搭配的是一件连衣裙。如图3所示,连衣裙在上装部位可以选择空or外套,在下装部位可以选择空or裤子,鞋和包部位是必须的。那么共有2X2=4种规则可以选择,即连衣裙+鞋+包的组合规则1、连衣裙+外套+鞋+包的组合规则2、连衣裙+裤子+鞋+包的组合规则3、连衣裙+外套+裤子+鞋+包的组合规则4。通过索引控制,在存储模块中可调出候选商品池中这4个部位的商品(外套、裤子、鞋和包),比如各类商品各200件,作为每类商品的数据对象集,用来生成搭配候选。
从候选商品池中调出某个类型的200件商品时,比如鞋,会按宝贝人气分+上架时间倒排,形成鞋类的数据对象集3,可将较热门和较新的商品优先参与目标商品(如连衣裙)组合。
接下来,可轮流使用图2所示的4个组合规则,产出数据对象组合。比如,首先使用“+鞋+包”规则(即组合规则1),各使用鞋和包队列中第一个商品,结合用户已有 的连衣裙商品,组成一个(连衣裙+鞋+包)3件套候选;然后,使用第二个“+外套+鞋+包”规则(即组合规则2),从外套队列中取第一个商品,从鞋和包的队列中各取第二个商品,组成(同样一个连衣裙但结合一个外套、新的鞋和包)4件套候选,传递到后续的模块。依次类推,当产出的候选满足数量要求(比如200个)时停止。
过滤部分
过滤部分可使用过滤规则来过滤掉不得体、在颜色纹理上不符合审美原则的数据对象组合。具体实施时,过滤规则可包括但不限于如下中的至少一种:针对款式的过滤规则、针对颜色的过滤规则、针对亮点的过滤规则等等。
具体实施时,针对款式的过滤规则可简单理解为图4所示表格中表征的款式搭配关系。参见图4所示的表格形式,由专家或达人沉淀的不能同时穿着在一身的款式组合组成,可生成2000或更多个格点;图4中仅示例性的展示了部分款式的搭配关系。格点中
和
表示可以同时穿在一身的款式,格点中
表示不适合同时穿在一身的款式,比如“毛衣”+“西装”、“T恤”+“羽绒裤”等。实质上,针对款式的过滤规则的作用是过滤掉在组成结构和季节温度上不符合得体要求的数据对象组合。
数据对象组合通常为3~5件套,对数据对象组合里面包含的所有商品进行两两查表,如果遇到不符合针对款式的过滤规则的组合,则该数据对象组合需删除。例如,某一数据对象组合包括:外套、上装及裤装;相应的,查询外套与上装是否符合针对款式的过滤规则,上装与裤装是否符合针对款式的过滤规则,以及外套与裤装是否符合针对款式的过滤规则。若外套和上装、上装与裤装、以及外套与裤装中有一个搭配组合通过查询属于不适合同时穿于一身的组合时,该数据对象组合需过滤掉,即从多个数据对象组合中删除。
通常每个商品的颜色信息会有1~3个颜色、纹理信息会含有1~3个纹理标签,如下:
商品A:
颜色信息:(Color1,AreaC1)、(Color2,AreaC2)…
纹理信息:(Texture1,AreaT1)、(Texture2,AreaT2)…
其中,Color是颜色标签,如胭脂红,相应的Area是该颜色标签在商品中的面积占比,如40%;类似的,Texture是纹理标签,如波点,相应的Area是该纹理标签在商品中的面积占比,如70%。
针对颜色的过滤规则可具体为:数据对象组合中主色超过3个需过滤掉。一个颜色 被定义为主色的条件是该颜色标签的面积Area>=30%,并且该颜色在HSV空间中的S>=0.14(否则是黑白灰色系,黑白灰不占用主色个数)。该规则的执行方式为,依次遍历一数据对象组合中的每个商品,并统计主色个数,当不同种类的主色个数大于3时,该数据对象组合需被过滤掉,即从多个数据对象组合中删除。
针对亮点的过滤规则可具体为:数据对象组合中亮点超过1个需过滤掉。一个颜色被定义为亮点的条件是,该颜色在HSV空间中的S>=0.72且V>=0.87(非常鲜艳);一个纹理被定义为亮点的条件是,该纹理的面积Area>=80%(大面积纹理)。该规则的执行方式为,依次遍历一数据对象组合中的每个商品,并统计亮点个数,当不同种类的亮点总数大于1时,该数据对象组合需被过滤掉,即从多个数据对象组合中删除。
经过上述过程后,剩余没有被过滤掉的数据对象组合即为能供用户参考的服饰搭配实例。
排序部分
通常经过上述过滤过程后,剩下一半左右可作为服饰搭配实例提供给用户。在提供给用户之前,可对这些服饰搭配实例进行排序,来决定最终面向用户的展现顺序。通常这个步骤是为了进一步优化搭配的点击率。具体实施,可根据各服饰搭配实例中每个商品的人气分(可以是销量、点击率、加购率等等,依据业务目标不同可以调整)计算搭配分,然后根据搭配分进行排序。然后按搭配分从高到低排序展现给用户。
图5示出了本申请另一实施例提供的数据对象处理方法的流程示意图。如图所示,所述数据对象处理方法包括:
201、获取与第一数据对象所属类型不同的至少一个类型的第二数据对象。
202、根据所述第一数据对象的属性及所述至少一个第二数据对象的属性,判定所述至少一个第二数据对象是否与所述第一数据对象搭配。
203、所述至少一个第二数据对象与所述第一数据对象搭配时,将所述至少一个第二数据对象提供给用户。
上述201中,至少一个类型的第二数据对象中各类型的第二数据对象可以在相应类型的数据对象集中获得。在服饰类电商场景下,上述第一数据对象、第二数据对象可以是服饰类商品,如连衣裙、外套、鞋、包等等。第一数据对象、第二数据对象可以是用于唯一标识商品的商品号、商品名称等等。
上述202中,若在服务类应用场景下,所述属性可包括如下至少一种:部位、款式、颜色信息、纹理信息、销售记录、加购率、点击率、上架时间。相应的,本步骤202可 包括如下至少一种实现步骤:
2021、所述第一数据对象的款式与至少一个第二数据对象中的任一第二数据对象的款式,判定所述第一数据对象及所述至少一个第二数据对象是否搭配;
2022、根据所述第一数据对象的颜色信息与至少一个第二数据对象的颜色信息确定主色的数量;根据所述主色的数量,判定所述第一数据对象及所述至少一个第二数据对象是否搭配;
2023、根据所述第一数据对象的颜色信息与所述至少一个第二数据对象的颜色信息确定是否具有颜色亮点作为第一确定结果;根据所述第一数据对象的纹理信息与所述至少一个第二数据对象的纹理信息确定是否具有纹理亮点作为第二确定结果,根据所述第一确定结果和所述第二确定结果,判定所述第一数据对象及所述至少一个第二数据对象是否搭配。
一个颜色被定义为亮点的条件是:该颜色在HSV空间中的S>=0.72且V>=0.87(非常鲜艳);一个纹理被定义为亮点的条件是,该纹理的面积Area>=80%(大面积纹理)。如果,根据所述第一确定结果和第二确定结果,得出第一数据对象与至少一个第二数据对象组合的亮点不大于1,则第一数据对象与至少一个第二数据对象搭配。
这里需要说明的是:上述2021~2023的具体实现可参见上述实施例中的相应内容,此处不再赘述。
本实施例提供的技术方案,从与第一数据对象所属类型不同的至少一个类型的数据对象集中提取用以与第一数据对象组合的第二数据对象,以生成多个数据对象组合,然后利用搭配规则,而非条件严苛的搭配模板,从多个数据对象组合中选出能供用户参考的服饰搭配实例,普适性强,在保证搭配质量的前提下产出的搭配更具丰富性;另外,通过效果实测,本申请实施例提供的技术方案能够为目标商品池99%的商品产出多套的搭配建议,并且人工评测满意率达到80%。
本申请各实施例提供的技术方案可应用在服饰搭配推荐场景中。例如,用户通过客户端应用(APP)挑选一条连衣裙,将挑选出的连衣裙加入购物车。客户端监听到该用户行为后,为用户推荐可与该连衣裙搭配的外套、鞋、包等等;并将这些能与连衣裙搭配的外套、鞋、包对应的页面信息作为推荐信息展示在用户界面上。例如,在一种可行的方案中,客户端针对该连衣裙向服务端发送搭配请求。服务端可采用上述各实施例提供的方法向客户端反馈能为用户推荐的外套、鞋、包等。即本实施例提供的技术方案还可采用如下硬件系统架构实现。如图6所示,服饰搭配信息的处理系统包括:
客户端301,用于响应于用户针对第一数据对象进行的操作,向服务端发送请求信息;接收所述服务端针对所述第一数据对象反馈的服饰搭配实例;将所述服饰搭配实例提供给所述用户;
服务端302,用于获取客户端针对所述第一数据对象发送的请求信息;从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中,提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;将所述多个数据对象组合中符合搭配规则的数据对象组合作为服饰搭配实例反馈至所述客户端。
本实施例提供的技术方案,从与第一数据对象所属类型不同的至少一个类型的数据对象集中提取用以与第一数据对象组合的第二数据对象,以生成多个数据对象组合,然后利用搭配规则,而非条件严苛的搭配模板,从多个数据对象组合中选出能供用户参考的服饰搭配实例,普适性强,在保证搭配质量的前提下产出的搭配更具丰富性;另外,通过效果实测,本申请实施例提供的技术方案能够为目标商品池99%的商品产出多套的搭配建议,并且人工评测满意率达到80%。
上述服饰搭配信息的处理系统中,各组成部分如客户端、服务端的具体实现方法将在如下各实施例中进行详细说明。
图7示出了本申请一实施例提供的所述服饰搭配信息的处理方法的流程示意图。本实施例提供的所述方法的执行主体可以是客户端,该客户端可以是手机、平板电脑、PDA(Personal Digital Assistant,个人数字助理)、POS(Point of Sales,销售终端)、车载电脑等任意终端设备。如图7所示,所述方法包括:
401、响应于用户针对第一数据对象进行的操作,向服务端发送请求信息。
402、接收所述服务端针对所述第一数据对象反馈的服饰搭配实例。
403、将所述服饰搭配实例提供给所述用户。
其中,所述服饰搭配实例是从多个数据对象组合中选出的符合搭配规则的数据对象组合;所述多个数据对象组合是将从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中提取的第二数据对象与所述第一数据对象组合生成的。
上述401中,指定操作可包括:用户点击相应控键的操作、用户发出针对第一数据对象的控制语音、或用户做出指定动作等,本实施例对此不作具体限定。
本实施例提供的技术方案,从与第一数据对象所属类型不同的至少一个类型的数据对象集中提取用以与第一数据对象组合的第二数据对象,以生成多个数据对象组合,然后利用搭配规则,而非条件严苛的搭配模板,从多个数据对象组合中选出能供用户参考 的服饰搭配实例,普适性强,在保证搭配质量的前提下产出的搭配更具丰富性;另外,通过效果实测,本申请实施例提供的技术方案能够为目标商品池99%的商品产出多套的搭配建议,并且人工评测满意率达到80%。
这里需要说明的是:有关服饰搭配实例生成过程可参见上述实施例中的相关内容,此处不再赘述。
进一步的,上述步骤403“将所述服饰搭配实例提供给所述用户”,可包括如下至少一种:
显示所述服饰搭配实例;
基于所述服饰搭配实例,在虚拟模特上展示虚拟试穿图像;
获取所述用户的照片,根据所述照片生成所述用户虚拟试穿的图像。
图8示出了本申请一实施例提供的服饰搭配信息的处理方法的流程示意图。本实施例提供的所述方法的执行主体可以是服务端,该服务端可以是普通的服务器、虚拟服务器或云端等,本实施例对此不作具体限定。如图8所示,所述方法包括:
501、获取客户端针对所述第一数据对象发送的请求信息。
502、从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中,提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合。
503、将所述多个数据对象组合中符合搭配规则的数据对象组合作为服饰搭配实例反馈至所述客户端。
上述步骤501~502的具体实现可参见上述实施例中的相关内容,此处不再赘述。
进一步的,本实施例提供的所述方法除上述步骤之外,还可实现上述各实施例中所描述的其他步骤,具体可参见前述各实施例中的描述。
本实施例提供的技术方案,从与第一数据对象所属类型不同的至少一个类型的数据对象集中提取用以与第一数据对象组合的第二数据对象,以生成多个数据对象组合,然后利用搭配规则,而非条件严苛的搭配模板,从多个数据对象组合中选出能供用户参考的服饰搭配实例,普适性强,在保证搭配质量的前提下产出的搭配更具丰富性;另外,通过效果实测,本申请实施例提供的技术方案能够为目标商品池99%的商品产出多套的搭配建议,并且人工评测满意率达到80%。
图9示出了本申请一实施例提供的服饰搭配信息的处理装置的结构示意图。如图9所示,所述数据对象处理装置包括:获取模块11、组合模块12及选择模块13。其中,所述获取模块11用于获取与第一数据对象所属类型不同的至少一个类型的数据对象集; 所述组合模块12用于从所述至少一个类型的数据对象集中提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;所述选择模块13用于从所述多个数据对象组合中,选出符合搭配规则的数据对象组合作为能供用户参考的服饰搭配实例。
本实施例提供的技术方案,从与第一数据对象所属类型不同的至少一个类型的数据对象集中提取用以与第一数据对象组合的第二数据对象,以生成多个数据对象组合,然后利用搭配规则,而非条件严苛的搭配模板,从多个数据对象组合中选出能供用户参考的服饰搭配实例,普适性强,在保证搭配质量的前提下产出的搭配更具丰富性;另外,通过效果实测,本申请实施例提供的技术方案能够为目标商品池99%的商品产出多套的搭配建议,并且人工评测满意率达到80%。
进一步的,所述组合模块12可具体用于:
获取至少一个组合规则;
使用所述至少一个组合规则,从所述至少一个类型的数据对象集中的部分数据对象集或全部数据对象集中分别选出一个数据对象作为第二数据对象与所述第一数据对象组合成数据对象组合;直至生成的不同数据对象组合的数量满足预置数量要求。
进一步的,所述至少一个组合规则包括:第一组合规则和第二组合规则;以及所述组合模块12还用于:
获取所述至少两个类型的数据对象集中各类型的数据对象集对应的当前排序序号Ni,其中,不同类型的数据对象集的i值不同;
从所述第一组合规则指定所有数据对象集中,分别获取排序为Ni的数据对象作为第二数据对象与所述第一数据对象组合成数据对象组合,并将所述第一组合规则指定的所有数据对象集各自对应的Ni更新为Ni+1;
从所述第二组合规则指定的所有数据对象集中,分别获取排序为Ni的数据对象作为第二数据对象与所述第一数据对象组合成数据对象组合,并将所述第二组合规则指定的所有数据对象集各自对应的Ni更新为Ni+1;
重复上述步骤,直至生成的不同组合实例的数量满足预置数量要求。
进一步的,所述选择模块13还用于:
获取所述多个数据对象组合中各数据对象组合所含数据对象的属性;
使用针对所述属性中至少一个属性项的过滤规则,判定所述多个数据对象组合中各数据对象组合所含数据对象的属性组合是否符合所述过滤规则;
将所述多个数据对象组合中所含数据对象的属性组合符合所述过滤规则的数据对象 组合删除。
进一步的,所述数据对象的属性包括如下至少一种属性项:类型、款式、颜色信息、纹理信息、销售记录、加购率、点击率、上架时间。
进一步的,过滤规则包括:针对款式的规则;相应的,所述选择模块13还用于:
获取所述第一数据对象组合所含的所述第一数据对象的款式以及与所述第一数据对象组合的至少一个第二数据对象的款式;
判定所述第一数据对象的款式分别与所述至少一个第二数据对象中各第二数据对象的款式组合是否符合所述针对款式的规则。
再进一步的,所述选择模块13还用于:所述第一数据对象的款式与所述至少一个第二数据对象中任一第二数据对象的款式组合符合所述针对款式的规则时,将所述第一数据对象组合删除。
进一步的,过滤规则包括:针对颜色的规则;以及所述选择模块13还用于:
获取所述第一数据对象组合所含的所述第一数据对象的颜色信息以及与所述第一数据对象组合的至少一个第二数据对象的颜色信息;
基于所述第一数据对象的颜色信息及所述至少一个第二数据对象的颜色信息,确定满足主色条件的主色数量;
判定所述主色数量是否符合所述针对颜色的规则。
再进一步的,所述选择模块还用于:所述主色数量大于第一阈值时,将所述第一数据对象组合删除。
进一步的,过滤规则包括:针对亮点的规则;以及所述选择模块13还用于:
获取所述第一数据对象组合所含的所述第一数据对象的颜色信息和纹理信息以及与所述第一数据对象组合的至少一个第二数据对象的颜色信息和纹理信息;
对所述第一数据对象的颜色信息和所述至少一个第二数据对象的颜色信息进行亮点分析,得到与亮点有关的第一分析结果;
对所述第一数据对象的纹理信息和所述至少一个搭配数据对象的纹理信息进行亮点分析,得到与亮点有关的第二分析结果;
判定所述第一分析结果和所述第二分析结果,是否符合所述针对亮点的规则。
再进一步的,所述选择模块13还用于:
根据所述第一分析结果和所述第二分析结果得出亮点数量大于第二阈值时,将所述第一数据对象组合删除。
进一步的,所述获取模块还用于:
获取所述第一类型数据对象池中各数据对象的属性;
根据所述第一类型数据对象池中各数据对象的属性,对所述第一类型数据对象池中的数据对象进行排序;
选取排序在前的多个数据对象组成所述第一类型的数据对象集。
进一步的,本实施例提供的所述方法还包括:排序模块。所述排序模块用于对多个服饰搭配实例进行排序;按照排序结果,将多个服饰搭配实例提供给用户。
这里需要说明的是:上述实施例提供的服饰搭配信息的处理装置可实现上述各方法实施例中描述的技术方案,上述各模块或单元具体实现的原理可参见上述各方法实施例中的相应内容,此处不再赘述。
图10示出了本申请一实施例提供的数据对象处理装置的结构示意图。如图10所示,所述数据对象处理装置包括:获取模块21、判定模块22及提供模块23。其中,所述获取模块21用于获取与第一数据对象所属类型不同的至少一个类型的第二数据对象;所述判定模块22用于根据所述第一数据对象的属性及所述至少一个第二数据对象的属性,判定所述至少一个第二数据对象是否与所述第一数据对象搭配;所述提供模块23用于所述至少一个第二数据对象与所述第一数据对象搭配时,将所述至少一个第二数据对象提供给用户。
本实施例提供的技术方案,从与第一数据对象所属类型不同的至少一个类型的数据对象集中提取用以与第一数据对象组合的第二数据对象,以生成多个数据对象组合,然后利用搭配规则,而非条件严苛的搭配模板,从多个数据对象组合中选出能供用户参考的服饰搭配实例,普适性强,在保证搭配质量的前提下产出的搭配更具丰富性;另外,通过效果实测,本申请实施例提供的技术方案能够为目标商品池99%的商品产出多套的搭配建议,并且人工评测满意率达到80%。
进一步的,在服饰类应用场景下,上述第一数据对象、第二数据对象即具体的服饰商品,相应的,上述第一数据对象、第二数据对象的属性可具体包括如下至少一种:部位、款式、颜色信息、纹理信息、销售记录、加购率、点击率、上架时间;相应的,所述判定模块22还用于:
所述第一数据对象的款式与至少一个第二数据对象中的任一第二数据对象的款式,判定所述第一数据对象及所述至少一个第二数据对象是否搭配;和/或
根据所述第一数据对象的颜色信息与至少一个第二数据对象的颜色信息确定主色的 数量;和/或
根据所述主色的数量,判定所述第一数据对象及所述至少一个第二数据对象是否搭配;和/或
根据所述第一数据对象的颜色信息与所述至少一个第二数据对象的颜色信息确定是否具有颜色亮点作为第一确定结果;根据所述第一数据对象的纹理信息与所述至少一个第二数据对象的纹理信息确定是否具有纹理亮点作为第二确定结果,根据所述第一确定结果和所述第二确定结果,判定所述第一数据对象及所述至少一个第二数据对象是否搭配。
这里需要说明的是:上述实施例提供的数据对象处理装置可实现上述各方法实施例中描述的技术方案,上述各模块或单元具体实现的原理可参见上述各方法实施例中的相应内容,此处不再赘述。
图11示出了本申请一实施例提供的服饰搭配信息的处理装置的结构示意图。如图11所示,所述数据对象处理装置包括:发送模块31、接收模块32及输出模块33。其中,发送模块31用于响应于用户针对第一数据对象进行的操作,向服务端发送请求信息;接收模块32用于接收所述服务端针对所述第一数据对象反馈的服饰搭配实例;所述输出模块33用于将所述服饰搭配实例提供给所述用户。其中,所述服饰搭配实例是从多个数据对象组合中选出的符合搭配规则的数据对象组合;所述多个数据对象组合是将从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中提取的第二数据对象与所述第一数据对象组合生成的。
进一步的,所述输出模块还用于:
显示所述服饰搭配实例;和/或
基于所述服饰搭配实例,在虚拟模特上展示虚拟试穿图像;和/或
获取所述用户的照片,根据所述照片生成所述用户虚拟试穿的图像。
本实施例提供的技术方案,从与第一数据对象所属类型不同的至少一个类型的数据对象集中提取用以与第一数据对象组合的第二数据对象,以生成多个数据对象组合,然后利用搭配规则,而非条件严苛的搭配模板,从多个数据对象组合中选出能供用户参考的服饰搭配实例,普适性强,在保证搭配质量的前提下产出的搭配更具丰富性;另外,通过效果实测,本申请实施例提供的技术方案能够为目标商品池99%的商品产出多套的搭配建议,并且人工评测满意率达到80%。
这里需要说明的是:上述实施例提供的服饰搭配信息的处理装置可实现上述各方法 实施例中描述的技术方案,上述各模块或单元具体实现的原理可参见上述各方法实施例中的相应内容,此处不再赘述。
图12示出了本申请一实施例提供的服饰搭配信息的处理装置的结构示意图。如图12所示,所述数据对象处理装置包括:获取模块41、提取模块42及反馈模块43。其中,所述获取模块41用于获取客户端针对所述第一数据对象发送的请求信息;所述提取模块42用于从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中,提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;所述反馈模块43用于将所述多个数据对象组合中符合搭配规则的数据对象组合作为服饰搭配实例反馈至所述客户端。
本实施例提供的技术方案,从与第一数据对象所属类型不同的至少一个类型的数据对象集中提取用以与第一数据对象组合的第二数据对象,以生成多个数据对象组合,然后利用搭配规则,而非条件严苛的搭配模板,从多个数据对象组合中选出能供用户参考的服饰搭配实例,普适性强,在保证搭配质量的前提下产出的搭配更具丰富性;另外,通过效果实测,本申请实施例提供的技术方案能够为目标商品池99%的商品产出多套的搭配建议,并且人工评测满意率达到80%。
这里需要说明的是:上述实施例提供的服饰搭配信息的处理装置可实现上述各方法实施例中描述的技术方案,上述各模块或单元具体实现的原理可参见上述各方法实施例中的相应内容,此处不再赘述。
图13示出了本申请一实施例提供的电子设备的结构示意图。该电子设备包括存储器51及处理器52。存储器51可被配置为存储其它各种数据对象以支持在电子设备上的操作。这些数据对象的示例包括用于在电子设备上操作的任何应用程序或方法的指令。存储器51可以由任何类型的易失性或非易失性存储设备或者它们的组合实现,如静态随机存取存储器(SRAM),电可擦除可编程只读存储器(EEPROM),可擦除可编程只读存储器(EPROM),可编程只读存储器(PROM),只读存储器(ROM),磁存储器,快闪存储器,磁盘或光盘。
所述处理器52,与所述存储器51耦合,用于执行所述存储器51中存储的所述程序,以用于:
获取与第一数据对象所属类型不同的至少一个类型的数据对象集;
从所述至少一个类型的数据对象集中提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;
从所述多个数据对象组合中,选出符合搭配规则的数据对象组合作为能供用户参考的服饰搭配实例。
其中,处理器52在执行存储器51中的程序时,除了上面的功能之外,还可实现其它功能,具体可参见前面各实施例的描述。
进一步的,如图13所示,电子设备还包括:显示器54、通信组件53、电源组件55、音频组件56等其它组件。图13中仅示意性给出部分组件,并不意味着电子设备只包括图13所示组件。
本申请一实施例还提供了一种电子设备。本实施例提供的所述电子设备的结构同上述电子设备实施例的结构类同,参见图13所示。该电子设备包括存储器及处理器。所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:
获取与第一数据对象所属类型不同的至少一个类型的第二数据对象;
根据所述第一数据对象的属性及所述至少一个第二数据对象的属性,判定所述至少一个第二数据对象是否与所述第一数据对象搭配;
所述至少一个第二数据对象与所述第一数据对象搭配时,将所述至少一个第二数据对象提供给用户。
其中,处理器在执行存储器中的程序时,除了上面的功能之外,还可实现其它功能,具体可参见前面各实施例的描述。
本申请一实施例还提供了一种客户端设备。本实施例提供的所述客户端设备的结构同上述电子设备实施例的结构类同,参见图13所示。所述客户端设备包括:存储器及处理器。其中,所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:
响应于用户针对第一数据对象进行的操作,向服务端发送请求信息;
接收所述服务端针对所述第一数据对象反馈的服饰搭配实例;
将所述服饰搭配实例提供给所述用户;
其中,所述服饰搭配实例是从多个数据对象组合中选出的符合搭配规则的数据对象组合;所述多个数据对象组合是将从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中提取的第二数据对象与所述第一数据对象组合生成的。
其中,处理器在执行存储器中的程序时,除了上面的功能之外,还可实现其它功能,具体可参见前面各实施例的描述。
本申请一实施例还提供了一种服务端设备。本实施例提供的所述服务端设备的结构同上述电子设备实施例的结构类同,参见图13所示。所述服务端设备包括:存储器及处理器。其中,所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:
获取客户端针对所述第一数据对象发送的请求信息;
从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中,提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;
将所述多个数据对象组合中符合搭配规则的数据对象组合作为服饰搭配实例反馈至所述客户端。
其中,处理器在执行存储器中的程序时,除了上面的功能之外,还可实现其它功能,具体可参见前面各实施例的描述。
相应地,本申请实施例还提供一种存储有计算机程序的计算机可读存储介质,所述计算机程序被计算机执行时能够实现上述各实施例提供的服饰搭配信息的处理方法的步骤或功能。
另外,本申请实施例还提供一种存储有计算机程序的计算机可读存储介质,所述计算机程序被计算机执行时能够实现上述各实施例提供的数据对象处理方法的步骤或功能。
以上所描述的装置实施例仅仅是示意性的,其中所述作为分离部件说明的单元可以是或者也可以不是物理上分开的,作为单元显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。本领域普通技术人员在不付出创造性的劳动的情况下,即可以理解并实施。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到各实施方式可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件。基于这样的理解,上述技术方案本质上或者说对现有技术做出贡献的部分可以以软件商品的形式体现出来,该计算机软件商品可以存储在计算机可读存储介质中,如ROM/RAM、磁碟、光盘等,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)执行各个实施例或者实施例的某些部分所述的方法。
最后应说明的是:以上实施例仅用以说明本申请的技术方案,而非对其限制;尽管参照前述实施例对本申请进行了详细的说明,本领域的普通技术人员应当理解:其依然 可以对前述各实施例所记载的技术方案进行修改,或者对其中部分技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围。
Claims (23)
- 一种服饰搭配信息的处理方法,其特征在于,包括:获取与第一数据对象所属类型不同的至少一个类型的数据对象集;从所述至少一个类型的数据对象集中提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;从所述多个数据对象组合中,选出符合搭配规则的数据对象组合作为供用户参考的服饰搭配实例。
- 根据权利要求1所述的方法,其特征在于,从所述至少一个类型的数据对象集中提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合,包括:获取至少一个组合规则;使用所述至少一个组合规则,从所述至少一个类型的数据对象集中的部分数据对象集或全部数据对象集中分别选出一个数据对象作为第二数据对象与所述第一数据对象组合成数据对象组合;直至生成的不同数据对象组合的数量满足预置数量要求。
- 根据权利要求2所述的方法,其特征在于,所述至少一个组合规则包括:第一组合规则和第二组合规则;以及使用所述第一组合规则和所述第二组合规则,从至少两个类型的数据对象集中的部分数据对象集或全部数据对象集中分别选出一个数据对象作为第二数据对象与所述第一数据对象组合成数据对象组合;直至生成的不同数据对象组合的数量满足预置数量要求,包括:获取所述至少两个类型的数据对象集中各类型的数据对象集对应的当前排序序号Ni,其中,不同类型的数据对象集的i值不同;从所述第一组合规则指定所有数据对象集中,分别获取排序为Ni的数据对象作为第二数据对象与所述第一数据对象组合成数据对象组合,并将所述第一组合规则指定的所有数据对象集各自对应的Ni更新为Ni+1;从所述第二组合规则指定的所有数据对象集中,分别获取排序为Ni的数据对象作为第二数据对象与所述第一数据对象组合成数据对象组合,并将所述第二组合规则指定的所有数据对象集各自对应的Ni更新为Ni+1;重复上述步骤,直至生成的不同组合实例的数量满足预置数量要求。
- 根据权利要求1至3中任一项所述的方法,其特征在于,从所述多个数据对象组合中,选出符合搭配规则的数据对象组合作为能供用户参考的服饰搭配实例,包括:获取所述多个数据对象组合中各数据对象组合所含数据对象的属性;使用针对所述属性中至少一个属性项的过滤规则,判定所述多个数据对象组合中各数据对象组合所含数据对象的属性组合是否符合所述过滤规则;将所述多个数据对象组合中所含数据对象的属性组合符合所述过滤规则的数据对象组合删除。
- 根据权利要求4所述的方法,其特征在于,所述数据对象的属性包括如下至少一种属性项:类型、款式、颜色信息、纹理信息、销售记录、加购率、点击率、上架时间。
- 根据权利要求5所述的方法,其特征在于,过滤规则包括:针对款式的规则;以及使用针对款式的规则,判定所述多个数据对象组合中的第一数据对象组合所含数据对象的属性组合是否符合所述针对款式的规则,包括:获取所述第一数据对象组合所含的所述第一数据对象的款式以及与所述第一数据对象组合的至少一个第二数据对象的款式;判定所述第一数据对象的款式分别与所述至少一个第二数据对象中各第二数据对象的款式组合是否符合所述针对款式的规则。
- 根据权利要求6所述的方法,其特征在于,将所述多个数据对象组合中所含数据对象的属性组合符合所述过滤规则的数据对象组合删除,包括:所述第一数据对象的款式与所述至少一个第二数据对象中任一第二数据对象的款式组合符合所述针对款式的规则时,将所述第一数据对象组合删除。
- 根据权利要求5所述的方法,其特征在于,过滤规则包括:针对颜色的规则;以及使用针对颜色的规则,判定所述多个数据对象组合中的第一数据对象组合所含数据对象的属性组合是否符合所述针对款式的规则,包括:获取所述第一数据对象组合所含的所述第一数据对象的颜色信息以及与所述第一数据对象组合的至少一个第二数据对象的颜色信息;基于所述第一数据对象的颜色信息及所述至少一个第二数据对象的颜色信息,确定满足主色条件的主色数量;判定所述主色数量是否符合所述针对颜色的规则。
- 根据权利要求8所述的方法,其特征在于,将所述多个数据对象组合中所含数据对象的属性组合符合所述过滤规则的数据对象组合删除,包括:所述主色数量大于第一阈值时,将所述第一数据对象组合删除。
- 根据权利要求5所述的方法,其特征在于,过滤规则包括:针对亮点的规则;以及使用针对亮点的规则,判定所述多个数据对象组合中的第一数据对象组合所含数据对象的属性组合是否符合所述针对亮点的规则,包括:获取所述第一数据对象组合所含的所述第一数据对象的颜色信息和纹理信息以及与所述第一数据对象组合的至少一个第二数据对象的颜色信息和纹理信息;对所述第一数据对象的颜色信息和所述至少一个第二数据对象的颜色信息进行亮点分析,得到与亮点有关的第一分析结果;对所述第一数据对象的纹理信息和所述至少一个搭配数据对象的纹理信息进行亮点分析,得到与亮点有关的第二分析结果;判定所述第一分析结果和所述第二分析结果,是否符合所述针对亮点的规则。
- 根据权利要求8所述的方法,其特征在于,将所述多个数据对象组合中所含数据对象的属性组合符合所述过滤规则的数据对象组合删除,包括:根据第一分析结果和第二分析结果得出亮点数量大于第二阈值时,将所述第一数据对象组合删除。
- 根据权利要求1至3中任一项所述的方法,其特征在于,获取与所述第一数据对象所属类型不同的第一类型的数据对象集,包括:获取所述第一类型数据对象池中各数据对象的属性;根据所述第一类型数据对象池中各数据对象的属性,对所述第一类型数据对象池中的数据对象进行排序;选取排序在前的多个数据对象组成所述第一类型的数据对象集。
- 根据权利要求1至3中任一项所述的方法,其特征在于,得到的能供用户参考的服饰搭配实例为多个的情况下,所述方法还包括:对多个服饰搭配实例进行排序;按照排序结果,将多个服饰搭配实例提供给用户。
- 一种数据对象处理方法,其特征在于,包括:获取与第一数据对象所属类型不同的至少一个类型的第二数据对象;根据所述第一数据对象的属性及所述至少一个第二数据对象的属性,判定所述至少一个第二数据对象是否与所述第一数据对象搭配;所述至少一个第二数据对象与所述第一数据对象搭配时,将所述至少一个第二数据对象提供给用户。
- 根据权利要求14所述的方法,其特征在于,所述属性包括如下至少一种:部位、款式、颜色信息、纹理信息、销售记录、加购率、点击率、上架时间;以及根据所述第一数据对象的属性及所述至少一个第二数据对象的属性,判定所述至少一个第二数据对象是否与所述第一数据对象搭配,包括如下至少一种:所述第一数据对象的款式与至少一个第二数据对象中的任一第二数据对象的款式,判定所述第一数据对象及所述至少一个第二数据对象是否搭配;根据所述第一数据对象的颜色信息与至少一个第二数据对象的颜色信息确定主色的数量;根据所述主色的数量,判定所述第一数据对象及所述至少一个第二数据对象是否搭配;根据所述第一数据对象的颜色信息与所述至少一个第二数据对象的颜色信息确定是否具有颜色亮点作为第一确定结果;根据所述第一数据对象的纹理信息与所述至少一个第二数据对象的纹理信息确定是否具有纹理亮点作为第二确定结果,根据所述第一确定结果和所述第二确定结果,判定所述第一数据对象及所述至少一个第二数据对象是否搭配。
- 一种服饰搭配信息的处理方法,其特征在于,包括:响应于用户针对第一数据对象进行的操作,向服务端发送请求信息;接收所述服务端针对所述第一数据对象反馈的服饰搭配实例;将所述服饰搭配实例提供给所述用户;其中,所述服饰搭配实例是从多个数据对象组合中选出的符合搭配规则的数据对象组合;所述多个数据对象组合是将从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中提取的第二数据对象与所述第一数据对象组合生成的。
- 根据权利要求16所述的方法,其特征在于,将所述服饰搭配实例提供给所述用户,包括如下至少一种:显示所述服饰搭配实例;基于所述服饰搭配实例,在虚拟模特上展示虚拟试穿图像;获取所述用户的照片,根据所述照片生成所述用户虚拟试穿的图像。
- 一种服饰搭配信息的处理方法,其特征在于,包括:获取客户端针对第一数据对象发送的请求信息;从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中,提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;将所述多个数据对象组合中符合搭配规则的数据对象组合作为服饰搭配实例反馈至所述客户端。
- 一种服饰搭配信息的处理系统,其特征在于,包括:客户端,用于响应于用户针对第一数据对象进行的操作,向服务端发送请求信息;接收所述服务端针对所述第一数据对象反馈的服饰搭配实例;将所述服饰搭配实例提供给所述用户;服务端,用于获取客户端针对所述第一数据对象发送的请求信息;从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中,提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;将所述多个数据对象组合中符合搭配规则的数据对象组合作为服饰搭配实例反馈至所述客户端。
- 一种电子设备,其特征在于,包括存储器及处理器;其中,所述存储器,用于存储程序;所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:获取与第一数据对象所属类型不同的至少一个类型的数据对象集;从所述至少一个类型的数据对象集中提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;从所述多个数据对象组合中,选出符合搭配规则的数据对象组合作为能供用户参考的服饰搭配实例。
- 一种电子设备,其特征在于,包括存储器及处理器;其中,所述存储器,用于存储程序;所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:获取与第一数据对象所属类型不同的至少一个类型的第二数据对象;根据所述第一数据对象的属性及所述至少一个第二数据对象的属性,判定所述至少一个第二数据对象是否与所述第一数据对象搭配;所述至少一个第二数据对象与所述第一数据对象搭配时,将所述至少一个第二数据对象提供给用户。
- 一种客户端设备,其特征在于,包括存储器及处理器;其中,所述存储器,用于存储程序;所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:响应于用户针对第一数据对象进行的操作,向服务端发送请求信息;接收所述服务端针对所述第一数据对象反馈的服饰搭配实例;将所述服饰搭配实例提供给所述用户;其中,所述服饰搭配实例是从多个数据对象组合中选出的符合搭配规则的数据对象组合;所述多个数据对象组合是将从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中提取的第二数据对象与所述第一数据对象组合生成的。
- 一种服务端设备,其特征在于,包括存储器及处理器;其中,所述存储器,用于存储程序;所述处理器,与所述存储器耦合,用于执行所述存储器中存储的所述程序,以用于:获取客户端针对第一数据对象发送的请求信息;从与所述第一数据对象所属类型不同的至少一个类型的数据对象集中,提取用以与所述第一数据对象组合的第二数据对象,以生成多个数据对象组合;将所述多个数据对象组合中符合搭配规则的数据对象组合作为服饰搭配实例反馈至所述客户端。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910101037.6 | 2019-01-31 | ||
| CN201910101037.6A CN111507790A (zh) | 2019-01-31 | 2019-01-31 | 服饰搭配信息的处理方法、数据对象处理方法、系统及设备 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2020156306A1 true WO2020156306A1 (zh) | 2020-08-06 |
Family
ID=71841163
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2020/073140 Ceased WO2020156306A1 (zh) | 2019-01-31 | 2020-01-20 | 服饰搭配信息的处理方法、数据对象处理方法、系统及设备 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111507790A (zh) |
| WO (1) | WO2020156306A1 (zh) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112488223A (zh) * | 2020-12-07 | 2021-03-12 | 平安消费金融有限公司 | 衣物收纳推荐方法、装置、终端设备及存储介质 |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104391971A (zh) * | 2014-12-05 | 2015-03-04 | 常州飞寻视讯信息科技有限公司 | 一种智能服饰搭配自动推荐方法 |
| CN104951966A (zh) * | 2015-07-13 | 2015-09-30 | 百度在线网络技术(北京)有限公司 | 推荐服饰商品的方法及装置 |
| CN106294420A (zh) * | 2015-05-25 | 2017-01-04 | 阿里巴巴集团控股有限公司 | 提供业务对象搭配信息的方法及装置 |
| CN106649383A (zh) * | 2015-11-02 | 2017-05-10 | 阿里巴巴集团控股有限公司 | 服饰管理方法及系统 |
| WO2017091220A1 (en) * | 2015-11-25 | 2017-06-01 | Murzin Yuri | Live dressing room |
| CN107248099A (zh) * | 2017-06-01 | 2017-10-13 | 深圳天珑无线科技有限公司 | 服饰搭配推荐方法、系统及存储介质 |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6083985B2 (ja) * | 2012-09-17 | 2017-02-22 | 株式会社カタボー | 服装コーディネート提案システム |
| US20140344102A1 (en) * | 2013-05-18 | 2014-11-20 | Chaya Cooper | Virtual Personal Shopping System |
| CN103400274B (zh) * | 2013-07-22 | 2016-06-08 | 郝芳莉 | 一种个性化服饰虚拟试穿戴方法 |
| CN106033547A (zh) * | 2015-03-12 | 2016-10-19 | 阿里巴巴集团控股有限公司 | 颜色搭配评价方法、服饰搭配推荐方法及装置 |
| CN107679155A (zh) * | 2017-09-27 | 2018-02-09 | 百度在线网络技术(北京)有限公司 | 服饰搭配库建立方法、信息推荐方法、装置、设备及介质 |
| CN108734557A (zh) * | 2018-05-18 | 2018-11-02 | 北京京东尚科信息技术有限公司 | 用于生成服饰推荐信息的方法、装置和系统 |
| CN108958062B (zh) * | 2018-08-09 | 2021-04-06 | 褚昱川 | 一种基于智能衣橱的衣物搭配方法及装置 |
-
2019
- 2019-01-31 CN CN201910101037.6A patent/CN111507790A/zh active Pending
-
2020
- 2020-01-20 WO PCT/CN2020/073140 patent/WO2020156306A1/zh not_active Ceased
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104391971A (zh) * | 2014-12-05 | 2015-03-04 | 常州飞寻视讯信息科技有限公司 | 一种智能服饰搭配自动推荐方法 |
| CN106294420A (zh) * | 2015-05-25 | 2017-01-04 | 阿里巴巴集团控股有限公司 | 提供业务对象搭配信息的方法及装置 |
| CN104951966A (zh) * | 2015-07-13 | 2015-09-30 | 百度在线网络技术(北京)有限公司 | 推荐服饰商品的方法及装置 |
| CN106649383A (zh) * | 2015-11-02 | 2017-05-10 | 阿里巴巴集团控股有限公司 | 服饰管理方法及系统 |
| WO2017091220A1 (en) * | 2015-11-25 | 2017-06-01 | Murzin Yuri | Live dressing room |
| CN107248099A (zh) * | 2017-06-01 | 2017-10-13 | 深圳天珑无线科技有限公司 | 服饰搭配推荐方法、系统及存储介质 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111507790A (zh) | 2020-08-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11062379B2 (en) | Automatic fashion outfit composition and recommendation system and method | |
| US10242396B2 (en) | Automatic color palette based recommendations for affiliated colors | |
| TWI760299B (zh) | 銷售物件資訊的圖片展示方法及裝置 | |
| US20200342320A1 (en) | Non-binary gender filter | |
| CN112015998B (zh) | 一种基于用户画像的商品推荐方法 | |
| US9679532B2 (en) | Automatic image-based recommendations using a color palette | |
| US9542704B2 (en) | Automatic image-based recommendations using a color palette | |
| US9524563B2 (en) | Automatic image-based recommendations using a color palette | |
| CN106294420B (zh) | 提供业务对象搭配信息的方法及装置 | |
| US20150095184A1 (en) | Recommending a personalized ensemble | |
| US20150379001A1 (en) | Automatic color validation of image metadata | |
| US20150379006A1 (en) | Automatic image-based recommendations using a color palette | |
| US10373231B2 (en) | System and method for social style mapping | |
| US9183510B1 (en) | Method and system for personalized recommendation of lifestyle items | |
| CN106649383A (zh) | 服饰管理方法及系统 | |
| CN105765574A (zh) | 用于物品和物品集合匹配的系统和方法 | |
| WO2020090054A1 (ja) | 情報処理システム、情報処理装置、サーバ装置、プログラム、又は方法 | |
| CN109597907A (zh) | 服饰管理方法和装置、电子设备、存储介质 | |
| US20230019794A1 (en) | Automatic item recommendations based on visual attributes and complementarity | |
| CN115905593A (zh) | 基于当季潮流风格对已有服装穿搭推荐的方法及系统 | |
| WO2020156306A1 (zh) | 服饰搭配信息的处理方法、数据对象处理方法、系统及设备 | |
| WO2020079235A1 (en) | Method and apparatus for accessing clothing | |
| WO2016188278A1 (zh) | 业务对象信息的收藏方法及装置 | |
| CN113496169B (zh) | 衣物穿搭的推荐方法、装置、设备及存储介质 | |
| CN118052612A (zh) | 直播信息处理方法、装置、设备与可读存储介质 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20748405 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 20748405 Country of ref document: EP Kind code of ref document: A1 |