Method, device, computer equipment and storage medium for obtaining jewelry characteristic information

Through multiple trained jewelry feature recognition models and neural networks, the problem of inconsistent and inefficient jewelry feature recognition in the existing technology is solved, and automated and standardized jewelry feature recognition is realized, which improves the recognition accuracy and efficiency.

CN113486905BActive Publication Date: 2024-12-03SHENZHEN XINGFANG TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110749060.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-07-01
Publication Date
2024-12-03
Estimated Expiration
2041-07-01

AI Technical Summary

Technical Problem

The existing jewelry feature recognition methods are inconsistent in the identification results due to differences in manual experience, which are inefficient and difficult to achieve standardized automatic recognition of jewelry features.

Method used

Using multiple trained jewelry feature recognition models, the jewelry image is recognized through neural networks, and the feature information under different attributes is determined, including mosaic, technology, style and material characteristics. The model parameters are adjusted using the detection network and semantic segmentation model to realize Automatic recognition.

Benefits of technology

It realizes automatic identification of jewelry features, consistent standards, improves identification accuracy and efficiency, and avoids errors caused by artificial experience differences.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113486905B_ABST
    Figure CN113486905B_ABST
Patent Text Reader

Abstract

The present application relates to a jewelry feature information acquisition method, device, computer equipment and storage medium, the method comprising: acquiring a jewelry image corresponding to a to-be-identified jewelry; inputting the jewelry image into a plurality of trained jewelry feature recognition models, so as to recognize the jewelry image through a neural network in each jewelry feature recognition model, and determine the jewelry feature information corresponding to the jewelry image; the plurality of jewelry feature recognition models are respectively used to recognize jewelry feature information under different jewelry attributes; according to the jewelry feature information output by each jewelry feature recognition model, the jewelry feature information corresponding to the to-be-identified jewelry is determined, thereby realizing automatic recognition of jewelry features, being able to recognize jewelry features of different jewelry in batches, and having consistent recognition standards, thereby avoiding recognition errors caused by experience differences, improving recognition accuracy while improving recognition speed, and effectively improving recognition efficiency of jewelry features.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device, computer equipment and storage medium for acquiring jewelry feature information. Background Art

[0002] With the continuous development of the jewelry industry and the continuous improvement of consumers' economic capabilities, the number of jewelry styles on the market is increasing.

[0003] In the prior art, before jewelry enters the circulation link, it is often necessary for specialized staff to identify the characteristics of the jewelry, such as setting corresponding labels for the jewelry after identifying the style of the jewelry, so as to facilitate subsequent sales. However, the existing manual identification method has the problem of inconsistent identification results due to differences in staff experience, and requires specialized training for staff, resulting in low identification efficiency. Summary of the invention

[0004] Based on this, it is necessary to provide a jewelry feature information acquisition method, device, computer equipment and storage medium to address the above technical problems.

[0005] A method for acquiring jewelry feature information, the method comprising:

[0006] Obtain jewelry images corresponding to the jewelry to be identified;

[0007] Input the jewelry image into a plurality of trained jewelry feature recognition models, so as to recognize the jewelry image through the neural network in each jewelry feature recognition model, and determine the jewelry feature information corresponding to the jewelry image; the plurality of jewelry feature recognition models are respectively used to recognize the jewelry feature information under different jewelry attributes;

[0008] According to the jewelry feature information output by each jewelry feature recognition model, the jewelry feature information corresponding to the jewelry to be recognized is determined.

[0009] In one embodiment, when the jewelry attributes include setting attributes, identifying the jewelry image through the neural network in each jewelry feature recognition model to determine the jewelry feature information corresponding to the jewelry image includes:

[0010] For each pixel point on the jewelry image, a detection frame position and a detection frame score corresponding to a detection frame centered on the pixel point are obtained through a neural network in a jewelry feature recognition model for identifying inlay properties; the detection frame score is used to characterize the probability that the area corresponding to the detection frame is a gemstone inlay area;

[0011] Get the target detection frame whose detection frame score exceeds the preset threshold;

[0012] Determining a gemstone inlay area in the jewelry image according to a detection frame position corresponding to the target detection frame;

[0013] The setting characteristics of the gemstones in the gemstone setting area are obtained.

[0014] In one embodiment, it also includes:

[0015] Acquire a training jewelry image as a training sample and an inlay feature label corresponding to the training jewelry image; the inlay feature label includes an inlay area label and an inlay method label;

[0016] Input the training jewelry image into the detection network model to be trained, so as to recognize the training jewelry image through the neural network in the detection network model, obtain the gemstone inlay area in the training jewelry image, and determine the inlay method information corresponding to the gemstone in the gemstone inlay area;

[0017] According to the gem setting area, setting method information and the setting feature label, the model parameters corresponding to the detection network model are adjusted, and the training jewelry image is input into the detection network model to be trained again until the training end condition is met, so as to obtain a jewelry feature recognition model for identifying setting attributes.

[0018] In one embodiment, when the jewelry attributes include craft attributes, identifying the jewelry image through the neural network in each jewelry feature recognition model to determine the jewelry feature information corresponding to the jewelry image includes:

[0019] Obtaining a feature vector corresponding to each pixel in the jewelry image through a neural network in a jewelry feature recognition model for identifying craft attributes; each component in the feature vector represents the probability of a corresponding craft feature;

[0020] According to each component in the feature vector, the process feature corresponding to each pixel point is determined;

[0021] The number of pixels corresponding to each process feature is obtained, and the process feature corresponding to the jewelry image is determined according to the number of pixels.

[0022] In one embodiment, it also includes:

[0023] Acquire a training jewelry image as a training sample and a process feature label corresponding to the training jewelry image;

[0024] Inputting the training jewelry image into the semantic segmentation model to be trained, so as to recognize the training jewelry image through the neural network recognition in the semantic segmentation model, and obtain the process features corresponding to the training jewelry image;

[0025] According to the process features and the process feature labels, the model parameters corresponding to the semantic segmentation model are adjusted, and the step of inputting the training jewelry image into the semantic segmentation model to be trained is executed again until the training end condition is met, so as to obtain a jewelry feature recognition model for identifying process attributes.

[0026] In one embodiment, identifying the jewelry image through the neural network in each jewelry feature recognition model to determine the jewelry feature information corresponding to the jewelry image includes:

[0027] Recognize the jewelry image through a neural network in a jewelry feature recognition model for identifying style attributes, and determine the style features corresponding to the jewelry image;

[0028] and / or,

[0029] The jewelry image is identified by a neural network in a jewelry feature recognition model for identifying material attributes, and material features corresponding to the jewelry image are determined.

[0030] In one embodiment, obtaining a jewelry image corresponding to the jewelry to be identified includes:

[0031] Obtain jewelry images corresponding to multiple angles of the jewelry to be identified;

[0032] The step of determining the jewelry feature information corresponding to the jewelry to be identified according to the jewelry feature information output by each jewelry feature recognition model includes:

[0033] For each jewelry feature information output by the jewelry feature recognition model, the accumulated quantity corresponding to each jewelry feature information is obtained, and the jewelry feature information whose accumulated quantity meets the preset quantity condition is used as the recognition result of the jewelry feature recognition model;

[0034] Based on the recognition results corresponding to each jewelry feature recognition model, the jewelry feature information corresponding to the jewelry to be recognized is determined.

[0035] A device for acquiring jewelry feature information, the device comprising:

[0036] A jewelry image acquisition module is used to acquire jewelry images corresponding to the jewelry to be identified;

[0037] A recognition module, used for inputting the jewelry image into a plurality of trained jewelry feature recognition models, so as to recognize the jewelry image through the neural network in each jewelry feature recognition model, and determine the jewelry feature information corresponding to the jewelry image; the plurality of jewelry feature recognition models are respectively used for recognizing the jewelry feature information under different jewelry attributes;

[0038] The jewelry feature determination module is used to determine the jewelry feature information corresponding to the jewelry to be identified according to the jewelry feature information output by each jewelry feature recognition model.

[0039] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the above methods when executing the computer program.

[0040] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of any of the above methods.

[0041] The above-mentioned jewelry feature information acquisition method, device, computer equipment and storage medium obtain jewelry images corresponding to the jewelry to be identified, input the jewelry images into multiple trained jewelry feature recognition models, and identify the jewelry images through the neural networks in each jewelry feature recognition model to determine the jewelry feature information corresponding to the jewelry image. Then, according to the jewelry feature information output by each jewelry feature recognition model, the jewelry feature information corresponding to the jewelry to be identified can be determined, thereby realizing automatic recognition of jewelry features, being able to batch recognize jewelry features of different jewelry, and having consistent recognition standards, thereby avoiding recognition errors caused by experience differences, improving recognition accuracy while increasing recognition speed, and effectively improving the recognition efficiency of jewelry features. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A diagram of an application environment of a jewelry feature information acquisition method in an embodiment;

[0043] Figure 2 It is a flowchart of a method for acquiring jewelry characteristic information in one embodiment;

[0044] Figure 3 It is a structural block diagram of a jewelry characteristic information acquisition device in one embodiment;

[0045] Figure 4 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0047] The present application provides a method for obtaining jewelry feature information, which can be applied to Figure 1In the application environment shown, the terminal 102 can communicate with the server 104 through the network, the terminal 102 can be but not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices, and the server 104 can be implemented as an independent server or a server cluster composed of multiple servers.

[0048] In one embodiment, Figure 2 As shown, a method for obtaining jewelry feature information is provided, and the method is applied to Figure 1 Taking the server 104 in the example as an example, the following steps may be included:

[0049] Step 201, obtaining a jewelry image corresponding to the jewelry to be identified.

[0050] As an example, the jewelry to be identified can be jewelry to be identified, which can be jewelry, handicrafts or other collections made from at least one of gold, silver and natural materials (minerals, rocks, organisms, etc.), or jade products.

[0051] The jewelry image may be an image of the jewelry to be identified, and at least a part of the jewelry image is an image used to describe the jewelry to be identified. For example, the jewelry image may be an image obtained by a user photographing the jewelry to be identified using a dedicated camera device, or may be obtained by photographing the jewelry to be identified using a camera function on a terminal device, such as photographing the jewelry to be identified using a camera function of a mobile terminal, recording a video, and processing the video to obtain the jewelry image.

[0052] In actual applications, the server can obtain the jewelry image corresponding to the jewelry to be identified. Specifically, the user can send a request for obtaining jewelry feature information corresponding to the jewelry to be identified to the server through the terminal, and send the jewelry image corresponding to the jewelry to be identified to the server, so that the server can obtain the jewelry image corresponding to the jewelry to be identified.

[0053] Step 202, inputting the jewelry image into a plurality of trained jewelry feature recognition models, so as to recognize the jewelry image through the neural network in each jewelry feature recognition model, and determine the jewelry feature information corresponding to the jewelry image;

[0054] As an example, multiple jewelry feature recognition models can be used to identify jewelry feature information under different jewelry attributes. Jewelry attributes can include at least two of the following: inlay attributes, craft attributes, style attributes, and material attributes. Among them, the inlay attribute can characterize whether the jewelry to be identified has inlaid gemstones and the inlay method of the gemstones; the craft attribute can characterize the processing method of the jewelry to be identified; the style attribute can characterize the style type or wearing method of the jewelry to be identified; the material attribute can characterize the material type of the jewelry to be identified.

[0055] In a specific implementation, after obtaining the jewelry image corresponding to the jewelry to be identified, the jewelry image can be input into multiple pre-trained jewelry feature recognition models. Through each jewelry feature recognition model, the input jewelry image is recognized respectively, and then the jewelry feature information corresponding to the jewelry image can be determined.

[0056] Step 203: Determine the jewelry feature information corresponding to the jewelry to be identified according to the jewelry feature information output by each jewelry feature recognition model.

[0057] After each jewelry feature recognition model recognizes the jewelry image and outputs the corresponding jewelry feature information, the server can determine the jewelry feature information corresponding to the jewelry to be recognized based on the jewelry feature information output by each jewelry feature recognition model. For example, the feature information output by each jewelry feature recognition model can be determined as the jewelry feature information corresponding to the jewelry to be recognized. For another example, the output jewelry feature information can be further screened, and the jewelry feature information corresponding to the jewelry to be recognized can be determined based on the screening result.

[0058] In this embodiment, by obtaining jewelry images corresponding to the jewelry to be identified, the jewelry images are input into multiple trained jewelry feature recognition models, so that the jewelry images are recognized through the neural networks in each jewelry feature recognition model, and the jewelry feature information corresponding to the jewelry images is determined. Then, according to the jewelry feature information output by each jewelry feature recognition model, the jewelry feature information corresponding to the jewelry to be identified can be determined, thereby realizing automatic recognition of jewelry features, being able to batch recognize jewelry features of different jewelry, and having consistent recognition standards, thereby avoiding recognition errors caused by experience differences, improving recognition accuracy while improving recognition speed, and effectively improving the recognition efficiency of jewelry features.

[0059] In one embodiment, when the jewelry attributes include setting attributes, identifying the jewelry image by using the neural network in each jewelry feature recognition model to determine the jewelry feature information corresponding to the jewelry image may include the following steps:

[0060] For each pixel point on the jewelry image, a detection frame position and a detection frame score corresponding to a detection frame centered on the pixel point are obtained through a neural network in a jewelry feature recognition model for identifying inlay properties; a target detection frame whose detection frame score exceeds a preset threshold is obtained; a gem inlay area in the jewelry image is determined according to the detection frame position corresponding to the target detection frame; and inlay features of gems in the gem inlay area are obtained.

[0061] As an example, the setting attribute may indicate whether the jewelry to be identified has a set of gemstones and the setting method of the gemstones. The jewelry feature information under the setting attribute may include at least one of the following:

[0062] There are set gemstones, there are no set gemstones, claw setting, bezel setting, wall setting, card setting, nail setting, common tooth setting, hidden setting, insert setting, and borderless setting.

[0063] Among them, the inlaid gemstones can include at least one of diamond, ruby, sapphire, emerald, cat's eye, jadeite, wood opal, tourmaline, garnet, pearl, crystal, agate, chalcedony, crushed disc, fluorite, hematite, malachite, sunstone, aquamarine, peridot, lapis lazuli, topaz, prehnite, amazonite, opal, spinel, chalcedony, jadeite, topaz, rhodonite, sky eye stone, Xiuyan jade, coral, pearl, amber, blue topaz, jasper, white jade, Shoushan stone, bloodstone, turquoise, sunstone, moonstone, and aventurine.

[0064] The detection frame position may include the height of the detection frame and the width of the detection frame. In practice, the detection frame position may be determined by the coordinates of the four corners of the detection frame. The detection frame score corresponding to the detection frame may be used to characterize the probability that the area corresponding to the detection frame is a gemstone inlay area;

[0065] In practical applications, when the jewelry attributes include setting attributes, after the jewelry image to be identified is input into multiple jewelry feature recognition models, the jewelry feature recognition model used to identify the setting attributes can identify the jewelry image through the neural network in the model.

[0066] When recognizing jewelry images, for each pixel point in the jewelry image, the neural network in the jewelry feature recognition model can obtain the detection frame position corresponding to the detection frame centered on the pixel point, and output the corresponding detection frame score for the detection frame.

[0067] After obtaining the detection frame score corresponding to the detection frame, the detection frame whose detection frame score exceeds the preset threshold can be determined and used as the target detection frame. For example, for each pixel in the jewelry image, the neural network can output a detection frame score in the range of [0,1], and then the detection frame whose detection frame score is higher than the preset threshold of 0.5 can be determined as the target detection frame.

[0068] After determining the target detection frame, the gem inlay area in the jewelry image can be determined according to the detection frame position corresponding to the target detection frame, and the inlay features of the gems in the gem inlay area can be obtained. Specifically, after determining the target detection frame, the neural network in the jewelry feature recognition model for identifying inlay properties can generate a feature vector for the target detection frame. The feature vector may include multiple components, and the number of components in the feature vector corresponds to the type of inlay method, and each component is the probability that the inlay method of the gem to be identified is a specific inlay method. For example, if the type of inlay method is N1, the feature vector includes N1 components, and the sum of the N1 components is equal to 1. After generating the feature vector corresponding to the target detection frame, the inlay method with the highest corresponding probability can be determined as the inlay feature of the gem in the identified gem inlay area.

[0069] In this embodiment, for each pixel point on the jewelry image, a detection frame position and a detection frame score corresponding to the detection frame centered on the pixel point are obtained through a neural network in a jewelry feature recognition model for identifying inlay properties, a target detection frame whose detection frame score exceeds a preset threshold is obtained, and according to the detection frame position corresponding to the target detection frame, the gemstone inlay area in the jewelry image is determined, and the inlay features of the gemstones in the gemstone inlay area are obtained, which can quickly identify the inlay method of the jewelry to be identified, and effectively improve the recognition efficiency of jewelry features.

[0070] In one embodiment, the method may further include the following steps:

[0071] The invention relates to a method for obtaining a training jewelry image as a training sample and an inlay feature label corresponding to the training jewelry image; inputting the training jewelry image into a detection network model to be trained, so as to identify the training jewelry image through a neural network in the detection network model, obtain a gemstone inlay area in the training jewelry image, and determine inlay method information corresponding to the gemstone in the gemstone inlay area; adjusting the model parameters corresponding to the detection network model according to the gemstone inlay area, inlay method information and the inlay feature label, and again executing the step of inputting the training jewelry image into the detection network model to be trained until a training end condition is met, thereby obtaining a jewelry feature recognition model for identifying inlay attributes.

[0072] As an example, the tessellation feature label includes a tessellation area label and a tessellation method label.

[0073] In a specific implementation, training jewelry images as training samples may be acquired in advance. In one example, the shooting conditions of the training jewelry images may be the same or similar to the shooting conditions corresponding to the jewelry to be identified. The jewelry corresponding to the training jewelry images may be a variety of different jewelry, such as the jewelry to be identified in the subsequent model use process.

[0074] After obtaining the training jewelry images, the training jewelry images can be manually annotated, and the inlay feature labels corresponding to each training jewelry image can be set. For example, a rectangular frame can be used to select the gemstones in the jewelry image to be trained, and the corresponding inlay area labels can be obtained. The inlay methods corresponding to the gemstones in the rectangular frame can be annotated to obtain inlay method labels. Then, the training jewelry images can be input into the detection network model to be trained, and the training jewelry images can be identified by the neural network in the detection network model, the gemstone inlay area in the training jewelry image can be obtained, and the inlay method information corresponding to the gemstones in the gemstone inlay area can be determined.

[0075] After determining the gem setting area and setting method information output by the detection network model, the model parameters corresponding to the detection network model can be adjusted according to the acquired gem setting area, setting method information and setting feature labels, and the step of inputting the training jewelry image into the detection network model to be trained is executed again. The above process is repeated until the training end condition is met. The current detection network model can be used as a jewelry feature recognition model for identifying setting attributes.

[0076] In this embodiment, by obtaining a training jewelry image as a training sample and an inlay feature label corresponding to the training jewelry image, the training jewelry image is input into the detection network model to be trained, the training jewelry image is recognized by the neural network in the detection network model, the gem inlay area in the training jewelry image is obtained, and the inlay method information corresponding to the gem in the gem inlay area is determined, and then the model parameters corresponding to the detection network model can be adjusted according to the gem inlay area, the inlay method information and the inlay feature label, and the training jewelry image is input into the detection network model to be trained again until the training end condition is met, so as to obtain a jewelry feature recognition model for identifying inlay attributes, and the jewelry feature recognition model for identifying inlay attributes can be trained, which provides a basis for subsequently improving the efficiency of jewelry feature recognition.

[0077] In one embodiment, when the jewelry attributes include craft attributes, identifying the jewelry image through the neural network in each jewelry feature recognition model to determine the jewelry feature information corresponding to the jewelry image may include the following steps:

[0078] The invention relates to a method for obtaining a feature vector corresponding to each pixel in the jewelry image through a neural network in a jewelry feature recognition model for identifying craft attributes; determining the craft feature corresponding to each pixel according to each component in the feature vector; obtaining the number of pixels corresponding to each craft feature, and determining the craft feature corresponding to the jewelry image according to the number of pixels.

[0079] As an example, the neural network in the jewelry feature recognition model for identifying craft attributes can be a semantic segmentation model. Each component in the feature vector can represent the probability of the corresponding craft feature. The craft features in the feature vector can include at least two of the following surface craft features: polished craft, wire drawing craft, frosted craft, and sanding craft.

[0080] In a specific implementation, when the jewelry attributes include craft attributes, after the jewelry image to be identified is input into multiple jewelry feature recognition models, the jewelry feature recognition model used to identify the craft attributes can identify the jewelry image through the neural network in the model to obtain the feature vector corresponding to each pixel in the jewelry image.

[0081] Specifically, when recognizing jewelry images, the neural network in the model can obtain the feature vector corresponding to each pixel in the jewelry image. For example, there are N2 kinds of jewelry craftsmanship. For the input jewelry image, the neural network can generate a feature map of the same size as the input image. Each pixel on the feature map corresponds to a vector containing N2 components. The sum of the N2 components is 1, and each component can represent the probability that the pixel corresponds to the corresponding craftsmanship.

[0082] After determining the feature vector corresponding to each pixel, for each pixel, the process feature corresponding to each pixel can be determined according to each component in the feature vector. Specifically, the process feature corresponding to the component with the largest value in the feature vector can be determined as the process feature of the corresponding pixel. In this way, the process feature corresponding to each pixel in the jewelry image can be determined.

[0083] After determining the process feature corresponding to each pixel point, the number of pixels corresponding to each process feature can be counted to obtain the number of pixels corresponding to each process feature, and the process feature corresponding to the jewelry image can be determined based on the number of pixels, such as the various identified process features, as the jewelry feature information of the jewelry to be identified, or one or more process features whose number of pixels exceeds a threshold (such as exceeding a preset number threshold) can be determined as the process feature corresponding to the jewelry image. By analyzing each pixel point in the jewelry image, multiple processing techniques applied to the jewelry to be identified can be quickly determined.

[0084] In this embodiment, a neural network in a jewelry feature recognition model for identifying process attributes is used to obtain a feature vector corresponding to each pixel in a jewelry image, and the process feature corresponding to each pixel is determined based on the components in the feature vector. The number of pixels corresponding to each process feature is obtained, and the process feature corresponding to the jewelry image is determined based on the number of pixels. This allows rapid identification of one or more process features of jewelry to be identified, thereby improving the recognition efficiency of multiple process features under process attributes.

[0085] In one embodiment, the method may further include the following steps:

[0086] A training jewelry image as a training sample and a process feature label corresponding to the training jewelry image are obtained; the training jewelry image is input into a semantic segmentation model to be trained, so as to recognize the training jewelry image through a neural network recognition in the semantic segmentation model, and obtain the process feature corresponding to the training jewelry image; according to the process feature and the process feature label, a model parameter corresponding to the semantic segmentation model is adjusted, and the step of inputting the training jewelry image into the semantic segmentation model to be trained is executed again until a training end condition is met, and a jewelry feature recognition model for identifying process attributes is obtained.

[0087] In practical applications, training jewelry images as training samples can be acquired in advance. Specifically, the shooting conditions of the training jewelry images can be the same or similar to the shooting conditions corresponding to the jewelry to be identified. The jewelry corresponding to the training jewelry images can be a variety of different jewelry, such as the jewelry to be identified in the subsequent model use process.

[0088] After obtaining the training jewelry images, the training jewelry images can be annotated and the process feature labels corresponding to each training jewelry image can be set. For example, one or more polygons can be used to select one or more regions with the same process treatment method in the training jewelry images, and the process feature labels corresponding to each region can be annotated.

[0089] After determining the process feature labels corresponding to the training jewelry images, the training jewelry images can be input into the semantic segmentation model to be trained, and the training jewelry images can be identified through the neural network in the semantic segmentation model to obtain the process features corresponding to the training jewelry images. Then, the model parameters corresponding to the detection network model can be adjusted according to the process features and process feature labels predicted for the training jewelry images, and the step of inputting the training jewelry images into the semantic segmentation model to be trained can be performed again. The above process is repeated until the training end condition is met, and the current semantic segmentation model can be used as a jewelry feature recognition model for identifying process attributes.

[0090] In this embodiment, by obtaining a training jewelry image as a training sample and a process feature label corresponding to the training jewelry image, the training jewelry image is input into the semantic segmentation model to be trained, the training jewelry image is recognized by the neural network recognition in the semantic segmentation model, the process feature corresponding to the training jewelry image is obtained, and the model parameters corresponding to the semantic segmentation model are adjusted according to the process feature and the process feature label, and the training jewelry image is input into the semantic segmentation model to be trained again until the training end condition is met, and a jewelry feature recognition model for identifying process attributes is obtained. The jewelry feature recognition model for identifying process attributes can be trained, which provides a basis for subsequently improving the efficiency of jewelry feature recognition.

[0091] In one embodiment, the identifying the jewelry image by using the neural network in each jewelry feature recognition model to determine the jewelry feature information corresponding to the jewelry image may also include:

[0092] The jewelry image is identified by a neural network in a jewelry feature recognition model for identifying style attributes to determine the style features corresponding to the jewelry image.

[0093] As an example, the style features corresponding to the style attribute may include at least two of the following: a ring, a pendant, a bracelet, a necklace, an earring, and an ear stud.

[0094] In practical applications, a jewelry feature recognition model for identifying style attributes can be pre-trained. Specifically, training jewelry images can be obtained, and the training jewelry images can be annotated, and style feature labels corresponding to each training jewelry image can be set, such as annotating each training jewelry image with a style feature label.

[0095] After determining the style feature label corresponding to the training jewelry image, the training jewelry image can be input into the classification network model to be trained, and the training jewelry image can be identified by the neural network in the classification network model to obtain the style features corresponding to the training jewelry image. Then, the model parameters corresponding to the classification network model can be adjusted according to the style features and style feature labels predicted for the training jewelry image, and the step of inputting the training jewelry image into the classification network model to be trained can be performed again. The above process is repeated until the training end condition is met, and the current classification network model can be used as a jewelry feature recognition model for identifying style attributes.

[0096] In the actual recognition process, the jewelry image can be recognized through the neural network in the jewelry feature recognition model used to recognize style attributes, and the style features corresponding to the jewelry image can be determined. Specifically, the jewelry feature recognition model used to recognize style attributes can be a classification network model. For each jewelry image input, the model can generate a feature vector through the neural network therein, and each component in the feature vector can represent the probability that the style of the jewelry to be recognized is the corresponding style feature. For example, the feature vector can be a vector including N3 components, the sum of the N3 components is 1, and each component can represent the probability that the style of the jewelry to be recognized is the corresponding style feature.

[0097] After obtaining the feature vector, the jewelry feature recognition model can determine the style feature with the highest probability as the style feature corresponding to the jewelry image based on the feature vector.

[0098] The step of identifying the jewelry image by using the neural network in each jewelry feature recognition model to determine the jewelry feature information corresponding to the jewelry image may also include the following steps:

[0099] The jewelry image is identified by a neural network in a jewelry feature recognition model for identifying material attributes, and material features corresponding to the jewelry image are determined.

[0100] As an example, the material feature corresponding to the material attribute may include a variety of metal materials used for processing jewelry, and specifically, for example, may include at least two of the following: gold material, silver material, and platinum material.

[0101] In a specific implementation, a jewelry feature recognition model for identifying material properties can be pre-trained. Specifically, training jewelry images can be obtained, and material feature labels corresponding to each training jewelry image can be annotated, such as setting a material feature label for each training jewelry image.

[0102] After determining the material feature labels corresponding to the training jewelry images, the training jewelry images can be input into the classification network model to be trained, and the training jewelry images can be identified by the neural network in the classification network model to obtain the material features corresponding to the training jewelry images. Then, the model parameters corresponding to the classification network model can be adjusted according to the material features and material feature labels predicted for the training jewelry images, and the step of inputting the training jewelry images into the classification network model to be trained can be performed again. The above process is repeated until the training end condition is met, and the current classification network model can be used as a jewelry feature recognition model for identifying material attributes.

[0103] In practical applications, the jewelry image can be identified through the neural network in the jewelry feature recognition model for identifying material attributes, and the style features corresponding to the jewelry image can be determined. Specifically, for each jewelry image input, the jewelry feature recognition model for identifying material attributes can generate a feature vector through the neural network therein, and each component in the feature vector can represent the probability that the material of the jewelry to be identified is the corresponding material feature. For example, the feature vector can be a vector including N4 components, the sum of the N4 components is 1, and each component can represent the probability that the material of the jewelry to be identified is the corresponding material feature. After obtaining the feature vector, the jewelry feature recognition model can determine the material feature with the highest probability as the material feature corresponding to the jewelry image based on the feature vector.

[0104] In this embodiment, the jewelry image can be identified by a neural network in a jewelry feature recognition model for identifying style attributes to determine the style features corresponding to the jewelry image, and / or the jewelry image can be identified by a neural network in a jewelry feature recognition model for identifying material attributes to determine the material features corresponding to the jewelry image, thereby quickly determining the style or material of the jewelry to be identified and effectively improving the recognition efficiency.

[0105] In one embodiment, the step of obtaining the jewelry image corresponding to the jewelry to be identified may include the following steps:

[0106] Obtain jewelry images corresponding to multiple angles of the jewelry to be identified,

[0107] In practical applications, jewelry images corresponding to multiple angles of the jewelry to be identified can be obtained. Specifically, the jewelry to be identified can be placed on a platform, and a uniform background can be used as the shooting background of the jewelry to be identified. Then, a dedicated camera device can be used to take a photo of the jewelry to be identified at a preset distance under stable external light conditions. During the photo shooting process, multiple jewelry images from different angles can be obtained by changing the placement of the jewelry to be identified.

[0108] Determining the jewelry feature information corresponding to the jewelry to be identified according to the jewelry feature information output by each jewelry feature recognition model may include the following steps:

[0109] For the jewelry feature information output by each jewelry feature recognition model, the cumulative quantity corresponding to each jewelry feature information is obtained, and the jewelry feature information whose cumulative quantity meets the preset quantity condition is used as the recognition result of the jewelry feature recognition model; based on the recognition results corresponding to each jewelry feature recognition model, the jewelry feature information corresponding to the jewelry to be recognized is determined.

[0110] In the specific implementation, since multiple jewelry images corresponding to the jewelry to be identified can be obtained, each jewelry image can obtain the corresponding jewelry feature information after being identified by the jewelry feature model, but the jewelry feature information corresponding to different jewelry images may be different. For example, for the jewelry A to be identified, M jewelry images (M≥2) are obtained, among which, for the material feature, the material feature identified by the jewelry image M1 is material feature Q1, and the material feature identified by the jewelry image M2 is material feature Q2, and there is a difference between the two.

[0111] Based on this, for each jewelry feature recognition model used to identify different jewelry attributes, the jewelry feature information output by it can be counted to obtain the cumulative number corresponding to each piece of jewelry feature information, and the jewelry feature information whose cumulative number meets the preset quantity condition can be used as the recognition result of the jewelry feature recognition model.

[0112] For example, for M jewelry images corresponding to the jewelry to be identified, after being input into the jewelry feature recognition model used to identify the inlay attributes, the model can respectively output M jewelry feature information about the inlay attributes, and then the inlay features with the highest cumulative number of the corresponding jewelry images can be determined as the jewelry feature information under the inlay attributes.

[0113] For another example, for the craft attributes, after inputting into the jewelry feature recognition model used to identify the craft attributes, the number of pixels corresponding to various craft features can be counted, and the accumulated number of pixels can be used as the accumulated number of corresponding craft features. Then, the craft features whose accumulated number exceeds the pixel number threshold can be determined as jewelry feature information under the craft attributes.

[0114] Alternatively, for style attributes, the style features with the highest cumulative number in the corresponding jewelry images can be determined as the jewelry feature information under the style attributes of the jewelry to be identified; correspondingly, for material attributes, the material features with the highest cumulative number in the corresponding jewelry images can be determined as the jewelry feature information under the material attributes of the jewelry to be identified.

[0115] After obtaining the recognition results corresponding to each jewelry feature recognition model, each recognition result can be used as jewelry feature information corresponding to the jewelry to be recognized.

[0116] In this embodiment, jewelry images corresponding to multiple angles of the jewelry to be identified can be obtained, and the cumulative number corresponding to each jewelry feature information output by each jewelry feature recognition model is obtained, and the jewelry feature information whose cumulative number meets the preset quantity condition is used as the recognition result of the jewelry feature recognition model. Then, based on the recognition results corresponding to each jewelry feature recognition model, the jewelry feature information corresponding to the jewelry to be identified can be determined. The final jewelry feature information corresponding to the jewelry to be identified can be determined based on the recognition results of multiple jewelry images. By integrating the information of multiple jewelry images, the accuracy of jewelry feature recognition is improved.

[0117] In one embodiment, when training a jewelry feature recognition model, after obtaining a training set including multiple training jewelry images, it can be divided according to a preset ratio, for example, randomly divided into two parts of 90% and 10%, where 90% of the training jewelry images can be used for model training and 10% of the training jewelry images can be used for model testing. For the training of jewelry feature recognition models for the same jewelry attribute, different recognition models can be trained separately by adjusting model parameters, and the effects of each recognition model can be compared based on the test data set, and the model with the highest recognition accuracy on the test data set is selected as the final jewelry feature recognition model.

[0118] It should be understood that although Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0119] In one embodiment, Figure 3 As shown, a device for acquiring jewelry feature information is provided, the device comprising:

[0120] The jewelry image acquisition module 301 is used to acquire the jewelry image corresponding to the jewelry to be identified;

[0121] The recognition module 302 is used to input the jewelry image into a plurality of trained jewelry feature recognition models, so as to recognize the jewelry image through the neural network in each jewelry feature recognition model, and determine the jewelry feature information corresponding to the jewelry image; the plurality of jewelry feature recognition models are respectively used to recognize the jewelry feature information under different jewelry attributes;

[0122] The jewelry feature determination module 303 is used to determine the jewelry feature information corresponding to the jewelry to be identified according to the jewelry feature information output by each jewelry feature recognition model.

[0123] In one embodiment, when the jewelry attribute includes a setting attribute, the identification module 302 includes:

[0124] A detection frame score acquisition submodule is used to obtain, for each pixel point on the jewelry image, a detection frame position and a detection frame score corresponding to the detection frame centered on the pixel point through a neural network in a jewelry feature recognition model used to identify inlay properties; the detection frame score is used to characterize the probability that the area corresponding to the detection frame is a gemstone inlay area;

[0125] The target detection frame determination submodule is used to obtain the target detection frame whose detection frame score exceeds a preset threshold;

[0126] An inlay area recognition submodule, used to determine the gemstone inlay area in the jewelry image according to the detection frame position corresponding to the target detection frame;

[0127] The inlay feature acquisition submodule is used to acquire the inlay features of the gemstones in the gemstone inlay area.

[0128] In one embodiment, it further includes:

[0129] An inlay feature label acquisition module is used to acquire a training jewelry image as a training sample and an inlay feature label corresponding to the training jewelry image; the inlay feature label includes an inlay area label and an inlay method label;

[0130] A first input module is used to input the training jewelry image into the detection network model to be trained, so as to recognize the training jewelry image through the neural network in the detection network model, obtain the gemstone inlay area in the training jewelry image, and determine the inlay method information corresponding to the gemstone in the gemstone inlay area;

[0131] The first parameter adjustment module is used to adjust the model parameters corresponding to the detection network model according to the gem setting area, the setting method information and the setting feature label, and to input the training jewelry image into the detection network model to be trained again until the training end condition is met, so as to obtain a jewelry feature recognition model for identifying setting attributes.

[0132] In one embodiment, when the jewelry attribute includes a craft attribute, the identification module 302 includes:

[0133] A feature vector determination submodule is used to obtain a feature vector corresponding to each pixel in the jewelry image through a neural network in a jewelry feature recognition model used to identify process attributes; each component in the feature vector represents the probability of a corresponding process feature;

[0134] A pixel process determination submodule is used to determine the process feature corresponding to each pixel point according to each component in the feature vector;

[0135] The process feature determination submodule is used to obtain the number of pixels corresponding to each process feature, and determine the process feature corresponding to the jewelry image according to the number of pixels.

[0136] In one embodiment, it further includes:

[0137] A process feature label acquisition module, used to acquire a training jewelry image as a training sample and a process feature label corresponding to the training jewelry image;

[0138] A second input module is used to input the training jewelry image into the semantic segmentation model to be trained, so as to identify the training jewelry image through the neural network recognition in the semantic segmentation model and obtain the process features corresponding to the training jewelry image;

[0139] The second parameter adjustment module is used to adjust the model parameters corresponding to the semantic segmentation model according to the process characteristics and the process feature labels, and to input the training jewelry image into the semantic segmentation model to be trained again until the training end condition is met, so as to obtain a jewelry feature recognition model for identifying process attributes.

[0140] In one embodiment, the identification module 302 includes:

[0141] A style feature recognition submodule, used to recognize the jewelry image through a neural network in a jewelry feature recognition model used to recognize style attributes, and determine the style features corresponding to the jewelry image;

[0142] and / or,

[0143] The material feature recognition submodule is used to recognize the jewelry image through a neural network in a jewelry feature recognition model used to recognize material attributes, and determine the material features corresponding to the jewelry image.

[0144] In one embodiment, the jewelry image acquisition module 301 is specifically used to acquire jewelry images corresponding to multiple angles of the jewelry to be identified;

[0145] The jewelry feature determination module 303 includes:

[0146] The recognition result statistics submodule is used to obtain the cumulative quantity corresponding to each piece of jewelry feature information output by each jewelry feature recognition model, and take the jewelry feature information whose cumulative quantity meets the preset quantity condition as the recognition result of the jewelry feature recognition model;

[0147] The recognition result fusion submodule is used to determine the jewelry feature information corresponding to the jewelry to be recognized based on the recognition results corresponding to each jewelry feature recognition model.

[0148] The specific definition of a jewelry feature information acquisition device can be found in the definition of a jewelry feature information acquisition method mentioned above, which will not be repeated here. Each module in the above-mentioned jewelry feature information acquisition device can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0149] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store jewelry images. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for obtaining jewelry feature information is implemented.

[0150] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0151] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0152] Obtain jewelry images corresponding to the jewelry to be identified;

[0153] Input the jewelry image into a plurality of trained jewelry feature recognition models, so as to recognize the jewelry image through the neural network in each jewelry feature recognition model, and determine the jewelry feature information corresponding to the jewelry image; the plurality of jewelry feature recognition models are respectively used to recognize the jewelry feature information under different jewelry attributes;

[0154] According to the jewelry feature information output by each jewelry feature recognition model, the jewelry feature information corresponding to the jewelry to be recognized is determined.

[0155] In one embodiment, when the processor executes the computer program, the steps in the other embodiments described above are also implemented.

[0156] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0157] Obtaining jewelry images corresponding to the jewelry to be identified;

[0158] Input the jewelry image into a plurality of trained jewelry feature recognition models, so as to recognize the jewelry image through the neural network in each jewelry feature recognition model, and determine the jewelry feature information corresponding to the jewelry image; the plurality of jewelry feature recognition models are respectively used to recognize the jewelry feature information under different jewelry attributes;

[0159] According to the jewelry feature information output by each jewelry feature recognition model, the jewelry feature information corresponding to the jewelry to be recognized is determined.

[0160] In one embodiment, when the computer program is executed by a processor, the steps in the other embodiments described above are also implemented.

[0161] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0162] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0163] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for acquiring jewelry feature information, characterized in that: The method comprises: Obtain jewelry images corresponding to the jewelry to be identified; Acquire multiple trained jewelry feature recognition models; the multiple jewelry feature recognition models are respectively used to recognize jewelry feature information under different jewelry attributes; the multiple jewelry feature recognition models include a jewelry feature recognition model for recognizing inlay attributes and a jewelry feature recognition model for recognizing craft attributes; The jewelry image is input into the jewelry feature recognition model for identifying inlay properties, and the jewelry feature recognition model for identifying inlay properties determines a target detection frame according to a detection frame score corresponding to each pixel point on the jewelry image, determines a gemstone inlay area in the jewelry image according to the target detection frame, and obtains the inlay features of the gemstones in the gemstone inlay area; wherein the jewelry feature recognition model for identifying inlay properties is trained based on the inlay feature labels corresponding to the training jewelry image, and the detection frame score is used to characterize the probability that the area corresponding to the detection frame is a gemstone inlay area; the inlay features characterize the jewelry inlay method; The jewelry image is input into the jewelry feature recognition model for identifying process attributes, and the jewelry feature recognition model for identifying process attributes obtains a feature vector corresponding to each pixel in the jewelry image, determines the process feature corresponding to each pixel according to each component in the feature vector, obtains the number of pixels corresponding to each process feature, and determines various process features whose number of pixels exceeds a threshold as the process feature corresponding to the jewelry image; each component in the feature vector represents the probability of the corresponding process feature; the process features corresponding to multiple components in the feature vector include polishing process, wire drawing process, frosting process, and sand pushing process; The jewelry feature information corresponding to the jewelry to be identified is determined according to the inlay feature and the process feature.

2. The method according to claim 1, characterized in that The step of determining the target detection frame according to the detection frame score corresponding to each pixel point on the jewelry image includes: For each pixel point on the jewelry image, a detection frame position and a detection frame score corresponding to a detection frame centered on the pixel point are obtained through a neural network in the jewelry feature recognition model for identifying inlay attributes; the detection frame score is used to characterize the probability that the area corresponding to the detection frame is a gemstone inlay area; Get the target detection frame whose detection frame score exceeds the preset threshold.

3. The method according to claim 1, characterized in that Also includes: Acquire a training jewelry image as a training sample and a mosaic feature label corresponding to the training jewelry image; The mosaic feature label includes a mosaic area label and a mosaic mode label; Input the training jewelry image into the detection network model to be trained, so as to recognize the training jewelry image through the neural network in the detection network model, obtain the gemstone inlay area in the training jewelry image, and determine the inlay method information corresponding to the gemstone in the gemstone inlay area; According to the gem setting area, setting method information and the setting feature label, the model parameters corresponding to the detection network model are adjusted, and the training jewelry image is input into the detection network model to be trained again until the training end condition is met, so as to obtain a jewelry feature recognition model for identifying setting attributes.

4. The method according to claim 1, characterized in that The inlay feature includes at least one of the following: Claw setting, bezel setting, wall setting, card setting, nail setting, common tooth setting, hidden setting, insert setting, and borderless setting.

5. The method according to claim 1, characterized in that Also includes: Acquire a training jewelry image as a training sample and a process feature label corresponding to the training jewelry image; Inputting the training jewelry image into the semantic segmentation model to be trained, so as to recognize the training jewelry image through the neural network recognition in the semantic segmentation model, and obtain the process features corresponding to the training jewelry image; According to the process features and the process feature labels, the model parameters corresponding to the semantic segmentation model are adjusted, and the step of inputting the training jewelry image into the semantic segmentation model to be trained is executed again until the training end condition is met, so as to obtain a jewelry feature recognition model for identifying process attributes.

6. The method according to claim 1, characterized in that Also includes: Recognize the jewelry image through a neural network in a jewelry feature recognition model for identifying style attributes, and determine the style features corresponding to the jewelry image; and / or, The jewelry image is identified by a neural network in a jewelry feature recognition model for identifying material attributes, and material features corresponding to the jewelry image are determined.

7. The method according to claim 1, characterized in that The step of obtaining a jewelry image corresponding to the jewelry to be identified includes: Obtain jewelry images corresponding to multiple angles of the jewelry to be identified; The step of determining the jewelry feature information corresponding to the jewelry to be identified according to the inlay feature and the process feature includes: For each jewelry feature information output by the jewelry feature recognition model, the accumulated quantity corresponding to each jewelry feature information is obtained, and the jewelry feature information whose accumulated quantity meets the preset quantity condition is used as the recognition result of the jewelry feature recognition model; the jewelry feature information output by the jewelry feature recognition model includes the inlay feature and the process feature; Based on the recognition results corresponding to each jewelry feature recognition model, the jewelry feature information corresponding to the jewelry to be recognized is determined.

8. A device for acquiring jewelry characteristic information, characterized in that: The device comprises: A jewelry image acquisition module is used to acquire jewelry images corresponding to the jewelry to be identified; The recognition module is used to obtain a plurality of trained jewelry feature recognition models, input the jewelry image into the jewelry feature recognition model for identifying inlay attributes, and the jewelry feature recognition model for identifying inlay attributes determines a target detection frame according to the detection frame score corresponding to each pixel point on the jewelry image, determines the gem inlay area in the jewelry image according to the target detection frame, and obtains the inlay features of the gems in the gem inlay area, wherein the inlay features represent the jewelry inlay method; input the jewelry image into the jewelry feature recognition model for identifying craft attributes, and the jewelry feature recognition model for identifying craft attributes obtains a feature vector corresponding to each pixel point in the jewelry image, determines the craft features corresponding to each pixel point according to each component in the feature vector, and obtains each craft feature. The number of pixels corresponding to the feature vector is determined, and various process features whose number of pixels exceeds a threshold are determined as process features corresponding to the jewelry image; each component in the feature vector represents the probability of the corresponding process feature; the process features corresponding to multiple components in the feature vector include polishing process, wire drawing process, frosting process, and sanding process; the multiple jewelry feature recognition models are respectively used to recognize jewelry feature information under different jewelry attributes, and the multiple jewelry feature recognition models include a jewelry feature recognition model for identifying inlay attributes and a jewelry feature recognition model for identifying process attributes; the jewelry feature recognition model for identifying inlay attributes is trained according to the inlay feature labels corresponding to the training jewelry image, and the detection frame score is used to represent the probability that the area corresponding to the detection frame is a gem inlay area; The jewelry feature determination module is used to determine the jewelry feature information corresponding to the jewelry to be identified according to the inlay feature and the process feature.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.