Method and system for evaluating attention and favorite degree of commodity package based on face data
Through the attention and liking prediction model based on facial data, the problems of low efficiency and poor standardization of packaging design in existing technologies are solved, and accurate consumer evaluation and personalized packaging design support are achieved.
Patent Information
- Application Number
- CN202510957972.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-11
- Publication Date
- 2025-10-21
AI Technical Summary
Existing technologies make it difficult to accurately evaluate consumers' attention and preference for product packaging, and ignore individual differences, resulting in low packaging design efficiency and poor standardization.
By acquiring packaging images and dividing them into regions, collecting users' eye movement data and facial expression data, establishing an attention and liking prediction model, and using convolutional neural networks and multi-layer perceptron networks for training, we can predict users' attention and liking for packaging features, and conduct personalized evaluation considering user identity information.
It achieves more accurate evaluation of packaging attention and preference, reduces the interference of language ability and social expectations, provides scientific design optimization and market feedback analysis, and supports customized marketing strategies.
Smart Images
Figure CN120823005A_ABST
Abstract
Description
Technical Field
[0001] The present invention mainly relates to the technical field of consumer behavior analysis, and in particular to a method and system for evaluating product packaging attention and liking based on facial data. Background Art
[0002] In the modern consumer market, product packaging not only serves the basic functions of conveying information and protecting products, but also becomes a crucial vehicle for attracting consumers' attention, stimulating purchasing desire, and conveying brand value. Especially in highly competitive industries such as fast-moving consumer goods and alcohol, subtle changes in packaging design often directly influence consumers' perceptions, preferences, and purchasing decisions.
[0003] Traditionally, the evaluation of product packaging effectiveness relies mainly on subjective methods such as questionnaires and interviews. These methods have many limitations: first, consumers' subjective expressions may be affected by factors such as language ability, willingness to express, and social expectations, and it is impossible to truly restore their subconscious perceptions and preferences; second, this type of evaluation method is time-consuming, inefficient, and poorly standardized, making it difficult to quickly iterate and guide packaging design.
[0004] In recent years, with the development of technologies such as artificial intelligence, computer vision, and human-computer interaction, perceptual research methods based on objective physiological or behavioral data have begun to emerge. Eye tracking technology, as a means of intuitively reflecting the distribution of users' visual attention, has been widely used in advertising, web interface, and product design research. Expression recognition technology can non-invasively capture consumers' emotional reactions to products, providing auxiliary evidence for packaging preference evaluation. However, most existing research remains at the stage of analyzing single-modal data, lacking systematic and generalizable integrated modeling methods, making it difficult to fully explore the synergistic information between multimodal data.
[0005] In addition, most current systems are unable to accurately assess individual differences and ignore the impact of identity background factors such as consumer gender, age, income level, and frequency of product use on packaging attention behavior and preference tendencies, making it difficult to serve the customized marketing and precise design of product packaging. Summary of the Invention
[0006] The technical problem to be solved by the present invention is to provide a method and system for evaluating the attention and preference of product packaging based on facial data, with the aim of accurately evaluating consumers' attention and preference for product packaging features, thereby providing guidance for product packaging.
[0007] The technical solution adopted by the present invention to solve the above technical problems is: In one aspect, the present invention provides a method for evaluating product packaging attention and likability based on facial data, the method comprising: Acquire a packaging image and divide the packaging image into regions according to regions where packaging features are located, wherein the packaging features include packaging color, logo, icon, and text; Collect eye movement data and facial expression data when users focus on different packaging areas; Analyze the user's attention to each area based on the collected eye movement data, and establish a region-eye movement data-attention dataset; Establish an attention prediction model, use the regions in the region-eye movement data-attention dataset as label data, concatenate them with the eye movement data as input, and use attention as output to train the established attention prediction model; Collect users' preferences for different packaging features and build a region-attention-facial expression-liking dataset; A likeability prediction model was established. The regions in the region-attention-facial expression-likeability dataset were used as label data, which were then concatenated with the eye movement data and facial expression data as input. The likeability was used as output to train the likeability prediction model. Predicting users' attention to different packaging areas based on the trained attention prediction model, and obtaining users' attention to packaging features in corresponding areas based on the users' attention to different areas; The trained liking prediction model predicts users' liking for different packaging areas based on the attention prediction results of the attention prediction model for different packaging areas and the users' facial expression data for different packaging areas, and obtains users' liking for the packaging features in the corresponding areas based on the users' liking for different areas.
[0008] Furthermore, dividing the packaging image into regions according to regions where packaging features are located includes dividing the packaging into regions according to the distribution of colors, logos, texts and icons on the packaging.
[0009] Furthermore, an attention prediction model is established based on convolutional neural networks; The training of the established attention prediction model includes: taking the region in the region-eye movement data-attention dataset as label data and splicing it with the eye movement data as the input feature of the convolutional neural network, taking the attention of the corresponding region as the true label, training the convolutional neural network, comparing the attention prediction result of the convolutional neural network with the true label during the training process, calculating the corresponding loss function, optimizing the convolutional neural network, and using the optimized convolutional neural network as the attention prediction model.
[0010] Furthermore, the analyzing of the user's attention to each area based on the collected eye movement data includes: generating an attention heat map based on the eye movement data of the user when paying attention to different packaging areas obtained by eye movement glasses, and dividing the attention into N levels according to the heat level of the attention heat map, where N is a positive integer greater than 1.
[0011] Furthermore, the eye movement data includes the user's gaze time, gaze frequency and saccade amplitude at different packaging areas.
[0012] Furthermore, a likeability prediction model was established based on convolutional neural networks and multi-layer perceptron networks; The training of the likeability prediction model specifically includes: Extract user emotional features from facial expressions in the region-attention-facial expression-likeness dataset based on a multi-layer perceptron network; The packaging area is concatenated with the attention and emotion features as input, and the user's likes are used as the true label. The convolutional neural network is trained. During the training process, the likes prediction results of the convolutional neural network are compared with the true label, and the corresponding loss function is calculated. The convolutional neural network is optimized and used as the likes prediction model.
[0013] Furthermore, the emotion characteristics include the user's emotion type and the duration of the corresponding emotion type.
[0014] Furthermore, the likeability acquisition in the region-attention-facial expression-likeability data set includes: dividing the user's likeability into M levels, where M is a positive integer greater than 1, obtaining the likeability evaluations of multiple users for the same packaging area, and taking the mode of the likeability evaluations of the multiple users as the likeability evaluation result of the corresponding packaging area, which is used to establish the region-attention-facial expression-likeability data set.
[0015] Furthermore, the method also includes: when establishing the attention prediction model and the liking prediction model, classifying users according to their identity information, and predicting the attention and liking of different types of users for different packaging features, wherein the user's identity information includes the user's age, gender, annual income range and frequency of product use.
[0016] On the other hand, the present invention also provides a product packaging attention and liking evaluation system based on facial data, the system comprising: Data collection module: used to create region-eye movement data-attention dataset and region-attention-facial expression-likability dataset; Model training module: establishes attention prediction model and likeability prediction model based on the region-eye movement data-attention data set and region-attention-facial expression-likeability data set; Attention Prediction Module: This module predicts the user's attention to the target product packaging features based on the packaging image area segmentation results and eye movement data; Likeability prediction module: predicts the user's likeability for the target product packaging features based on the attention prediction results of the attention prediction model and the user's facial expression data for the corresponding area of the target product.
[0017] Beneficial effects of the present invention: (1) Establish an attention prediction model through eye movement data, and predict the user's attention to different product packaging features based on the user's eye movement data based on the attention prediction model. Establish a product packaging feature preference prediction model based on the user's facial expression data and attention when paying attention to the product packaging to predict the user's preference for different packaging features. This can effectively avoid the interference of factors such as the user's language ability, expression willingness and social expectations on the evaluation of packaging attention and preference, evaluate the user's perception and preference for different product packaging based on physiological parameters, and provide more intuitive and reliable scientific and efficient technical support for product packaging design optimization and market feedback analysis; (2) The purpose of using attention obtained from eye movement data and facial expression data to collaboratively establish a likeability prediction model is to obtain more accurate likeability prediction results and reduce the misjudgment of likeability prediction based on single data; (3) In the process of establishing the attention prediction model and the liking prediction model, the attention and liking of different types of users towards product packaging are predicted based on the users' age, gender, annual income range and product usage frequency, so as to formulate different product packaging strategies for different types of target users based on the attention and liking prediction results. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 This is a flow chart of the method for evaluating product packaging attention and likability based on facial data according to the present invention; Figure 2 This is a schematic diagram of the structure of the product packaging attention and liking evaluation system based on facial data according to the present invention. DETAILED DESCRIPTION
[0019] The present invention discloses a method and system for evaluating product packaging attention and liking based on facial data. The method first divides the packaging area according to packaging features, collects eye movement data of users when paying attention to different packaging areas and facial expression data that appears simultaneously with the eye movement data, analyzes the user's attention when paying attention to different product packaging features based on the eye movement data, and trains an attention prediction model based on the eye movement data and attention. Subsequently, when predicting the user's attention to a target product, the user's attention to different packaging features can be obtained from the attention prediction model based on the user's eye movement data. At the same time, a liking prediction model is established based on the user's attention to different packaging areas, expression data, and liking evaluation results for product packaging features. The user's liking for different packaging features is obtained based on the trained liking prediction model, thereby realizing joint modeling and personalized prediction of product packaging attention and liking, thereby providing support for product packaging design optimization.
[0020] like Figure 1 As shown, the method for evaluating product packaging attention and liking based on facial data of the present invention includes the following contents: Step S1: Acquire a packaging image and divide the packaging image into regions according to regions where packaging features are located.
[0021] Since the actual purpose of attention prediction and preference prediction is to judge the attention and preference of consumers for packaging features such as color, logo, text and icon on the packaging, when performing area division, the packaging area can be divided specifically according to the distribution of color, logo, text and icon on the product packaging.
[0022] Step S2: Collect eye movement data and facial expression data when the user focuses on different packaging areas.
[0023] Eye movement data collection: Testers obtain eye movement data by wearing eye-tracking glasses to observe the target product. The eye movement data collected by the eye-tracking glasses include the testers' gaze time, number of gazes, and eye saccade amplitude on different packaging features of the product.
[0024] Facial Expression Data Acquisition: Cameras on the target product display shelves capture facial images that correspond in time with the eye movement data. These images are then imported into Imotions software for analysis to determine the test subject's emotional signature, including emotion type and duration. Emotion types include happiness, sadness, surprise, fear, disgust, and more. Emotion duration is determined by the duration of periods such as a raised mouth corner or a frown.
[0025] Step S3: Analyze the user's attention to each packaging area based on the collected eye movement data, and establish a region-eye movement data-attention dataset.
[0026] Attention acquisition: The testers' eye movement data is put into Imotions software for analysis to obtain the attention heat map of each product packaging feature. The attention heat map is quantified and the attention is divided into N levels according to the heat level after quantification of the attention heat map, where N is a positive integer greater than 1.
[0027] The eye movement data, the packaging area corresponding to the eye movement data, and the quantified attention are matched to establish a region-eye movement data-attention dataset.
[0028] Step S4: Collect users' preferences for different packaging features and establish a region-attention-facial expression-liking dataset.
[0029] Obtaining Likeability: Conduct a questionnaire survey on consumers to obtain their likeability evaluation results for different product packaging areas. Divide the user's likeability into M levels, where M is a positive integer greater than 1. For example, the likeability evaluation results are divided into levels of 1 to 10 points. The higher the level, the higher the user's likeability.
[0030] The likeability evaluations of multiple users for the same packaging area are obtained, and the mode of the likeability evaluations of multiple users is used as the likeability evaluation result of the corresponding packaging area to establish the area-attention-facial expression-likeability dataset.
[0031] Preferably, in order to predict the attention and liking of different types of users towards product packaging features, so as to formulate different product packaging strategies for different target users based on the attention and liking prediction results, in the data collection stage of establishing the region-eye movement data-attention dataset and the region-attention-facial expression-liking dataset, the acquired eye movement data, facial expression data, and the corresponding attention and liking should cover testers of different genders, different age groups, different annual incomes and different product usage frequencies, and the identity information of different testers should be taken into consideration in the subsequent training of the attention prediction model and the liking prediction model.
[0032] Step S5: Establish an attention prediction model, use the region in the region-eye movement data-attention dataset as label data and splice it with the eye movement data as input, and use the attention as output to train the established attention prediction model.
[0033] In this embodiment, an attention prediction model is established based on a convolutional neural network. The region in the region-eye movement data-attention dataset is used as label data and spliced with the eye movement data as the input feature of the convolutional neural network. The attention of the corresponding region is used as the true label, and the convolutional neural network is trained. During the training process, the attention prediction results of the convolutional neural network are compared with the true label, the corresponding loss function is calculated, the convolutional neural network is optimized, and the optimized convolutional neural network is used as the attention prediction model.
[0034] Step S6: Establish a likeability prediction model, use the regions in the region-attention-facial expression-likeability dataset as label data, concatenate them with the eye movement data and facial expression data as input, and use likeability as output to train the likeability prediction model.
[0035] In this embodiment, a likeability prediction model is established based on a multi-layer perceptron network and a convolutional neural network.
[0036] Firstly, the user's emotional features are extracted from facial expressions in the region-attention-facial expression-likeness dataset based on a multi-layer perceptron network.
[0037] The packaging area is then concatenated with the attention and emotional features as the input of the convolutional neural network, and the user's likes are used as the true label to train the convolutional neural network. During the training process, the likes prediction results of the convolutional neural network are compared with the true label, and the corresponding loss function is calculated. The convolutional neural network is optimized and the optimized convolutional neural network is used as the likes prediction model.
[0038] Step S7: predicting the user's attention to different packaging areas based on the trained attention prediction model, and obtaining the user's attention to the packaging features of the corresponding areas based on the user's attention to different areas.
[0039] Step S8: The trained liking prediction model predicts the user's liking for different packaging areas based on the attention prediction results of the attention prediction model for different packaging areas and the user's facial expression data for different packaging areas, and obtains the user's liking for the packaging features of the corresponding areas based on the user's liking for different areas.
[0040] In addition, in the present application, in the process of establishing the attention prediction model and the likeability prediction model, according to the attention and likeability prediction requirements, the user's eye movement data and facial expression data of the entire package can also be used to model and train the attention prediction model and the likeability prediction model to achieve the prediction of the attention and likeability of the entire package.
[0041] The present invention also provides a product packaging attention and liking evaluation system based on facial data, the system is as follows Figure 2 As shown, it includes data collection module, model training module, attention prediction module and likeability prediction module: Data collection module: used to create region-eye movement data-attention dataset and region-attention-facial expression-likability dataset; Model training module: establishes attention prediction model and likeability prediction model based on the region-eye movement data-attention data set and region-attention-facial expression-likeability data set; Attention Prediction Module: This module predicts the user's attention to the target product packaging features based on the packaging image area segmentation results and eye movement data; Likeability prediction module: predicts the user's likeability for the target product packaging features based on the attention prediction results of the attention prediction model and the user's facial expression data for the corresponding area of the target product.
Claims
1. A method for evaluating product packaging attention and likability based on facial data, characterized in that: The method comprises: Obtaining a packaging image, and dividing the packaging image into regions according to regions where packaging features are located; Collect eye movement data and facial expression data when users focus on different packaging areas; Analyze the user's attention to each area based on the collected eye movement data, and establish a region-eye movement data-attention dataset; Establish an attention prediction model, use the regions in the region-eye movement data-attention dataset as label data, concatenate them with the eye movement data as input, and use attention as output to train the established attention prediction model; Collect users' preferences for different packaging features and build a region-attention-facial expression-liking dataset; A likeability prediction model was established. The regions in the region-attention-facial expression-likeability dataset were used as label data, which were then concatenated with the eye movement data and facial expression data as input. The likeability was used as output to train the likeability prediction model. Predicting users' attention to different packaging areas based on the trained attention prediction model, and obtaining users' attention to packaging features in corresponding areas based on the users' attention to different areas; The trained liking prediction model predicts users' liking for different packaging areas based on the attention prediction results of the attention prediction model for different packaging areas and the users' facial expression data for different packaging areas, and obtains users' liking for the packaging features in the corresponding areas based on the users' liking for different areas.
2. The method for evaluating product packaging attention and liking based on facial data according to claim 1, characterized in that: The dividing the packaging image into regions according to the regions where the packaging features are located includes dividing the packaging into regions according to the distribution of colors, logos, texts and icons on the packaging.
3. The method for evaluating product packaging attention and liking based on facial data according to claim 1, characterized in that: Establish an attention prediction model based on convolutional neural network; The training of the established attention prediction model includes: taking the region in the region-eye movement data-attention data set as label data and splicing it with the eye movement data as the input feature of the convolutional neural network, taking the attention of the corresponding region as the true label, training the convolutional neural network, comparing the attention prediction result of the convolutional neural network with the true label during the training process, calculating the corresponding loss function, optimizing the convolutional neural network, and using the optimized convolutional neural network as the attention prediction model.
4. The method for evaluating product packaging attention and liking based on facial data according to claim 1, characterized in that: The analyzing of the user's attention to each area based on the collected eye movement data includes: generating an attention heat map based on the eye movement data of the user when paying attention to different packaging areas obtained by eye movement glasses, and dividing the attention into N levels according to the heat level of the attention heat map, where N is a positive integer greater than 1.
5. The method for evaluating product packaging attention and liking based on facial data according to claim 1, characterized in that: The eye movement data includes the user's gaze time, gaze times and saccade amplitude at different packaging areas.
6. The method for evaluating product packaging attention and liking based on facial data according to claim 1, characterized in that: Establish a likeability prediction model based on convolutional neural networks and multi-layer perceptron networks; The training of the likeability prediction model specifically includes: Extract user emotional features from facial expressions in the region-attention-facial expression-likeness dataset based on a multi-layer perceptron network; The packaging area is concatenated with the attention and emotion features as input, and the user's likes are used as the true label. The convolutional neural network is trained. During the training process, the likes prediction results of the convolutional neural network are compared with the true label, and the corresponding loss function is calculated. The convolutional neural network is optimized and used as the likes prediction model.
7. The method for evaluating packaging attention and liking based on facial expression data according to claim 6, wherein: The emotion characteristics include the user's emotion type and the duration of the corresponding emotion type.
8. The method for evaluating product packaging attention and liking based on facial data according to claim 6, characterized in that: The likeability acquisition in the region-attention-facial expression-likeability dataset includes: dividing the user's likeability into M levels, where M is a positive integer greater than 1, obtaining multiple users' likeability evaluations for the same packaging region, and taking the mode of the likeability evaluations of the multiple users as the likeability evaluation result of the corresponding packaging region for establishing the region-attention-facial expression-likeability dataset.
9. The method for evaluating product packaging attention and liking based on facial data according to claim 1, characterized in that: The method further includes: when establishing the attention prediction model and the preference prediction model, classifying users according to their identity information, and predicting the attention and preference of different types of users for different packaging features. The user identity information includes the user's age, gender, annual income range, and frequency of product use.
10. A product packaging attention and likability evaluation system based on facial data, used to implement the product packaging attention and likability evaluation method based on facial data according to any one of claims 1 to 9, characterized in that: The system comprises: Data collection module: used to create region-eye movement data-attention dataset and region-attention-facial expression-likability dataset; Model training module: establishes attention prediction model and likeability prediction model based on the region-eye movement data-attention data set and region-attention-facial expression-likeability data set; Attention Prediction Module: This module predicts the user's attention to the target product packaging features based on the packaging image area segmentation results and eye movement data; Likeability prediction module: predicts the user's likeability for the target product packaging features based on the attention prediction results of the attention prediction model and the user's facial expression data for the corresponding area of the target product.