Motion brave measurement method based on 3D-CNN and Transform

By combining 3D-CNN and Transformer neural networks and utilizing motion video data and kinematic parameters, the problem of low reliability of self-report questionnaire assessment results is solved, achieving accurate and objective measurement of exercise courage with a prediction accuracy of 92%.

CN118887698AInactive Publication Date: 2024-11-01NORTHEASTERN UNIV AT QINHUANGDAO
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410844874.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-27
Publication Date
2024-11-01
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing technologies rely on self-report questionnaires when assessing athletic courage, which are subject to social expectation bias, resulting in low reliability of assessment results. Furthermore, they do not delve into the specific behavioral performance of athletes during exercise.

Method used

We employ a neural network based on 3D-CNN and Transformer, combined with motion video data, to identify and evaluate the level of courage in motion through convolutional operations and self-attention mechanisms. We also combine kinematic parameters to construct a dataset and train a model to accurately measure five courage dimensions and the overall level.

Benefits of technology

It has achieved accurate measurement of athletic courage, improved the objectivity and reliability of the assessment results, and the model prediction accuracy has reached over 85%, especially the overall level prediction accuracy has reached 92%.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118887698A_ABST
    Figure CN118887698A_ABST
Patent Text Reader

Abstract

The invention discloses a 3D-CNN and Transform-based movement brave measurement method, and relates to the technical field of movement psychology and computer vision, the method utilizes advantage complementation of convolution operation and a self-attention mechanism to enhance representation learning, and the method has time and space modeling ability, local and global modeling ability, contour modeling ability and key point modeling ability. The implementation of the method comprises the steps of SCS-31 measurement of the brave level of movement, movement data collection, data set establishment, model construction and training evaluation. According to the method, the problem of insufficient accuracy of a traditional method in movement brave measurement is solved, the effect of accurately measuring the movement brave is achieved, and the method has the theoretical value of deeply understanding the relationship between the movement brave and the movement performance. The method is wide in application prospect and plays an important role in research in the field of exercise psychology and improvement of psychological quality of athletes.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of sports psychology, and in particular to a sports courage measurement method based on 3D-CNN and Transformer. Background Art

[0002] Sports courage generally refers to the bravery and determination shown when engaging in sports activities or sports competitions. It involves the ability to face challenges, overcome difficulties, take risks, and remain calm and focused under pressure. This kind of courage is considered a key factor in sports performance because it can help athletes effectively manage and overcome anxiety, tension, stress, and fear. In sports, sports courage is often used to enhance athletes' coping abilities, thereby achieving ideal sports performance and increasing the chances of winning. In recent years, with the development of psychology, research in the field of sports psychology has also received increasing attention.

[0003] As an important part of society, college students are of special importance in sports psychology research. Past studies have usually relied on self-report questionnaires to assess individuals' sports courage levels, but this method is often interfered by social desirability bias and default response sets, which affects the reliability of the evaluation results. Therefore, scholars have begun to call for the introduction of objective behavioral measurement methods in sports psychology research. With the help of sensors, artificial intelligence and other technologies, recent research on sports psychology has begun to use objective measurement methods for evaluation.

[0004] Vaulting is a jumping sport in traditional gymnastics. It has high requirements on the speed, jumping, balance and psychological quality of athletes. Sports courage is a key factor. Previous studies have shown that there is a significant positive correlation between sports courage and sports performance. However, most current studies only focus on the relationship between indirect variables and sports courage, and have not conducted in-depth research on the specific movements and behaviors of athletes during exercise.

[0005] In order to measure sports courage more accurately and improve the reliability of evaluation results, this paper proposes to use deep learning technology, combined with a neural network fused with 3D-CNN and Transformer, to identify and evaluate sports courage levels from vaulting horse sports videos. This method can automatically extract key features during the movement, deeply study the relationship between athletes' sports performance and courage levels, and provide important value for sports psychology research. Summary of the invention

[0006] This paper successfully developed a sports courage measurement method based on 3D-CNN and Transformer, which uses the complementary advantages of convolution operation and self-attention mechanism to accurately identify the sports courage level of individuals during vaulting and detect significant differences in kinematic parameters between different subjects. Through detailed analysis of the performance and courage scores of athletes and non-athletes, the reasons for the differences are systematically described, and the performance differences of individuals with different courage levels in five dimensions are deeply explored, providing a theoretical explanation for the relationship between courage and performance.

[0007] The invention can accurately measure five dimensions of sports courage: Mastery (MT), Determination (DT), Assertiveness (AT), Venturesome (VS), and Self-Sacrifice (SB) levels as well as the overall level. On the data set constructed by the invention, the prediction accuracy of the network model of the invention for the five dimensions exceeded 85%, and the prediction accuracy for the overall level of sports courage reached 92%.

[0008] In view of the defects in the prior art, the present invention provides a method for measuring sports courage based on 3D-CNN and Transformer, which includes:

[0009] S1: SCS-31 measures the level of sports courage. College student volunteers were recruited to complete the questionnaire survey and their sports courage level was assessed using the SCS-31. Participants with a history of mental illness, diabetes, etc. were excluded, and the collected data were screened and cleaned, including excluding outliers and cleaning data according to criteria such as answer logic and speed. Finally, descriptive statistical results were summarized, including the mean, standard deviation, and the ratio of women to athletes in the sample.

[0010] S2: Motion data collection. Videos of volunteers performing vaulting were recorded, and YOLOV8 was used to perform semantic segmentation of the human body to obtain key frames. The key frames were preprocessed by cropping, cutting out, aligning, and other steps to ensure that the person was in the center of the image. From these preprocessed key frames, ten key frames representing sub-processes such as run-up, take-off, crossing the horse's back, falling, and landing were randomly selected. Next, YOLOV8-pose was used to extract skeletal key point information from these key frames. These key point information is stored in the form of (xk, yk) coordinates, where k represents the kth key point. In order to reduce the interference of small offsets of key points to the network, this method uses a Gaussian function to generate a key point heat map, and all points around the key points are considered as key points. The parameter σ of the Gaussian function controls the variance of the generated Gaussian map, thereby controlling the smoothness of the heat map. The calculation formula of the Gaussian heat map is The generated key point heat map presents a Gaussian probability distribution, where the grayscale value of the true target point approaches the maximum value of 1, while the value around the target point gradually decreases with the increase of distance. All heat maps are superimposed along the time dimension to obtain the 3D heat map volume of the key point, whose size is K×T×H×W, where K is the number of key points, T is the time dimension, and H and W are the height and width of the frame. This method not only reduces the amount of data, but also improves the accuracy of action recognition. Combined with the human silhouette image, we obtain two basic modalities: the human silhouette image and the 3D heat map volume of the key point. The data of these two modalities complement each other and provide information for the subsequent sports courage measurement.

[0011] S3: Establish a data set. The collected athlete images are processed, including extracting key frames, calculating the center of gravity information and joint angle information, and generating a 2D energy map. The center of gravity information, joint angle information and body swing area in the vaulting horse movement are counted and recorded in a table. Then, the relationship between different dimensions (MT, DT, AT, VS, SB) and the center of gravity information, joint angle information and body swing area in the vaulting horse movement is analyzed, five of which are data measured by S1. Finally, the center of gravity and joint angle data in the vaulting horse process are combined with the courage information.

[0012] S4: Model construction, training and evaluation:

[0013] S401: A 3D data structure model that integrates 3D-CNN and Two-Stage-Transformer in time and space is constructed, and a routing mechanism is added. The routing mechanism is introduced to reduce the computational complexity. The routing mechanism can dynamically select the appropriate calculation path according to different input conditions, thereby improving the efficiency and performance of the model.

[0014] S402: Using the cross attention mechanism to fuse the features of the two data modalities of human silhouette and key points as the input of the model. Through this mechanism, the model can automatically learn and pay attention to the relevant information between different modalities, thereby improving the representation ability and robustness of human posture. ; Input the input data to the 3D-CNN branch and the Two-Stage-Transformer branch for processing;

[0015] S403: Calculate the accuracy and loss when using 3D-CNN and Two-Stage-Transformer separately; and finally calculate the accuracy and loss when using Two-Stage-Transformer which is a fusion of 3D-CNN and Two-Stage-Transformer. Train the model and select the combined loss to train the network, including: Fusion loss (Lfuse): used to constrain the consistency and accuracy of the model during feature fusion. Contour loss (Lsil): used to constrain the model to accurately capture the human body contour information and the sensitivity to morphological changes during movement. Key point loss (Ljoi): used to constrain the model to accurately predict the key point information of the human body and the sensitivity to posture changes. The total loss is defined as L=w1×Lfuse+w2×Lsil+w3×Ljoi. The output result is the binary classification result of the total courage level;

[0016] S404: Figure 6 As shown, the confusion matrix is ​​observed and the classification accuracy of different dimensions (MT, DT, AT, VS, SB) is analyzed. The heat map is used to display the key areas that the model focuses on during the vaulting horse movement (the posture areas of the neck, back, arms and thighs). The experimental results are analyzed and the performance indicators of the model are evaluated, including accuracy, precision, recall rate and F1 score.

[0017] On the basis of the above technical solution, compared with the prior art, the advantages of the present invention are:

[0018] 1. Comprehensive research method: This invention adopts a comprehensive research method, combining deep learning technology and traditional psychological scale evaluation, as well as kinematic parameter analysis of vaulting horse technology, to comprehensively evaluate the level of sports courage from multiple perspectives.

[0019] 2. Objective measurement: An objective method for measuring sports courage was introduced, which not only used traditional self-report questionnaires but also combined vaulting horse video data and deep learning technology to improve the objectivity and reliability of the evaluation results.

[0020] 3. Application of deep learning technology: The model is able to simultaneously focus on the temporal relationship between different frames and the spatial relationship within the same frame during the vaulting process. The organic combination of convolution operation and self-attention mechanism enables the model to simultaneously capture the local features and global dependencies of the two motion modes during the vaulting process. The use of human silhouettes retains the body contour information, while key points retain the internal structure information of the body. These two modes complement each other and can more comprehensively depict the vaulting process. Comprehensive consideration is given to three aspects: time and space, local and global, and contour and key points. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1.Overall flow chart: provides an overall summary of the operation of the entire method.

[0022] Figure 2 .Flowchart of motion data collection.

[0023] Figure 3 .Flowchart of model construction and training evaluation.

[0024] Figure 4 The SCS scale measures the quantitative information and courage level of the test subjects in sports courage.

[0025] Figure 5 .Detailed schematic diagram of the human silhouette and key point heat map extraction method in motion data collection. Part (a) shows the whole process of modal extraction. Part (b) shows the process of constructing the key point 3D heat map volume.

[0026] Figure 6 . Specific schematic diagram of model training evaluation: Part (a) shows the results of the ablation experiment, from top to bottom are the training curves of 3D-CNN, Two-Stage-Transformer and the fusion of 3D-CNN and Two-Stage-Transformer. Part (b) shows the confusion matrix of the five dimensions of SCS and the total courage level predicted by motion modality. Part (c) shows the area that the model focuses on. The color depth of the heat map indicates the contribution of the corresponding area to the model classification decision. The dark area indicates that the model pays more attention to this area and plays a more important role in classification.

[0027] Figure 7 . Detection results of vaulting horse motion features. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0029] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

[0030] like Figure 1 As shown in the figure, a sports courage measurement method based on 3D-CNN and Transformer is used to quantitatively evaluate the five dimensions and overall level of sports courage, including the following steps:

[0031] S1: The embodiment of the present invention discloses a method for measuring the level of sports courage by SCS-31;

[0032] 1. Select college students and conduct sample screening, and use SCS-31 to assess the participants' sports courage level;

[0033] 2. Patients with a history of mental illness, diabetes, etc. were excluded;

[0034] 3. Screen the collected data and exclude univariate and multivariate outliers;

[0035] 4. Clean the data according to answer logic, answer speed and other criteria;

[0036] 5. Such as Figure 4 As shown, the descriptive statistics are summarized, including the mean, standard deviation, and the ratio of females to athletes in the sample distribution, of which 23% are females and 37% are athletes. In the entire sample, the calculated range of sports courage is 97 to 149 points, with a mean of 114.32 points and a standard deviation of 12.65;

[0037] In the step S1, the SCS-31 scale is a sports courage scale structure test questionnaire containing five factors. The scale is divided into five dimensions: Mastery (MT), Determination (DT), Assertiveness (AT), Venturesome (VS) and Self-Sacrifice (SB). Mastery (MT) includes 7 items; Determination (DT) includes 9 items; Assertiveness (AT) includes 7 items; Venturesome (VS) includes 4 items; Self-Sacrifice (SB) includes 4 items. They form a total of 31 items (including 7 reverse scoring items). SCS adopts a 5-point Likert scale, where 1 = "completely disagree" and 5 = "completely agree". The higher the score of each dimension, the better the athlete's sports courage in this aspect. The higher the total score of the questionnaire, the better the overall sports courage of the athlete. The scoring standard of SCS is directly related to the level of sports courage.

[0038] S2: Figure 2 As shown, an embodiment of the present invention discloses a method for collecting motion data, using YOLOV8 to perform semantic segmentation of human body in a vaulting horse video to obtain key frames;

[0039] 1. Perform pre-processing such as cropping, cutting out, and aligning on the extracted key frames to ensure that the person is located in the center of the image;

[0040] 2. Randomly select ten key frames representing the sub-processes of running up, taking off, crossing the horse's back, falling and landing from the pre-processed key frames;

[0041] 3. Use YOLOV8-pose to extract key point information and generate key point heat map to effectively reduce the amount of data and improve courage recognition accuracy;

[0042] 4. Such as Figure 4 As shown, the key point heat map is combined with the human silhouette map to obtain two basic modalities: human silhouette map and 3D heat map volume of key points;

[0043] 5. The data from the two modalities complement each other to more comprehensively characterize the vaulting process and provide a rich information basis for in-depth analysis.

[0044] In step S2, the use of YOLOV8_pose is further divided into the following steps:

[0045] 1. Apply YOLOV8-pose to each key frame to extract the skeleton key point information of the human body. These key point information are stored in the form of (xk, yk) coordinates, where k represents the kth key point. 2. In order to reduce the interference of small offsets of key points on the network, the Gaussian function is used to generate the key point heat map, and the points around the key points are considered as key points. 3. The parameter σ of the Gaussian function controls the variance of the generated Gaussian map, thereby controlling the smoothness of the heat map. 4. The generated key point heat map presents a Gaussian probability distribution. The grayscale value of the true target point approaches the maximum value of 1, while the value around the target point gradually decreases with the increase of the distance to the calibrated key point. 5. Superimpose all heat maps along the time dimension to obtain the 3D heat map volume of the key point, whose size is K x T x H x W, where K is the number of key points, T is the time dimension, and H and W are the height and width of the frame. 6. Calculation formula of Gaussian heat map

[0046] S3: The embodiment of the present invention provides a method for establishing a data set to analyze the correlation between kinematic parameters and sports courage.

[0047] 1. Processing athlete images:

[0048] Extract keyframes: Select keyframes from the video to represent different stages in the vaulting process.

[0049] Calculate center of mass information: Use visual methods to calculate the center of mass position of the athlete in each keyframe.

[0050] Calculate joint angle information: Use image processing technology to obtain the angle information of each joint of the athlete in each key frame.

[0051] Generate 2D energy map: Stack the silhouette sequence during the vaulting process into a 2D energy map to more intuitively display the athlete's movement characteristics.

[0052] 2. Statistical motion feature information, such as Figure 7 As shown:

[0053] Statistical center of gravity information: Record the center of gravity position of the athlete in each key frame, including changes in horizontal and vertical directions.

[0054] Statistical joint angle information: record the angle changes of each joint of the athlete in each key frame, such as the angles of the elbow, shoulder, hip and knee joints.

[0055] Record in a table: Organize and record the center of gravity information and joint angle information obtained in a table to facilitate subsequent analysis and comparison.

[0056] 3. Analyze the relationship between different dimensions and motion characteristics:

[0057] The differences in movement characteristics among individuals with different courage dimensions (MT, DT, AT, VS, SB) were compared, such as the different manifestations of center of gravity changes, joint angle changes, and body swing areas.

[0058] To analyze the differences in movement characteristics of individuals with different courage levels in vaulting horse sports, and explore their possible causes and influencing factors.

[0059] 4. Combine the center of gravity, joint angle data and courage information for correlation analysis:

[0060] The center of gravity and joint angle data were combined with the information of each courage dimension to analyze the correlation between them.

[0061] Through correlation analysis, the sport characteristics of individuals with different courage levels in vaulting horse sports, as well as the degree of correlation between courage and sports performance, were explored.

[0062] 5. Determine the characteristics of individuals with different courage levels in vaulting:

[0063] Based on the analysis results, the differences in movement characteristics such as center of gravity changes, joint angle changes, and body swing areas in vaulting horse sports among individuals with different courage levels were determined.

[0064] This paper summarizes the differences in movement characteristics of individuals with different courage levels and the impact of these differences on vaulting horse performance, providing a theoretical basis and practical guidance for further research.

[0065] By comprehensively analyzing five dimensions, it was found that compared with individuals with low courage levels, individuals with high courage levels had smaller elbow and shoulder joint angles during vaulting, larger hip and knee joint angles, and the standard deviation of horizontal change of center of gravity of individuals with high courage levels was significantly lower than that of individuals with low courage levels.

[0066] like Figure 3 As shown, the embodiment of the present invention discloses a method for model construction and training evaluation, constructs a model that adapts to 3D data structure and provides a model training method;

[0067] S401: A 3D data structure model that integrates 3D-CNN and Two-Stage-Transformer in time and space is constructed, and a routing mechanism is added. The routing mechanism is introduced to reduce the computational complexity. The routing mechanism can dynamically select the appropriate calculation path according to different input conditions, thereby improving the efficiency and performance of the model.

[0068] S402: Using the cross attention mechanism to fuse the features of the two data modalities of human silhouette and key points as the input of the model. Through this mechanism, the model can automatically learn and pay attention to the relevant information between different modalities, thereby improving the representation ability and robustness of human posture. ; Input the input data to the 3D-CNN branch and the Two-Stage-Transformer branch for processing;

[0069] S403: Calculate the accuracy and loss when using 3D-CNN and Two-Stage-Transformer separately; and finally calculate the accuracy and loss when using Two-Stage-Transformer which is a fusion of 3D-CNN and Two-Stage-Transformer. Train the model and select the combined loss to train the network, including: Fusion loss (Lfuse): used to constrain the consistency and accuracy of the model during feature fusion. Contour loss (Lsil): used to constrain the model to accurately capture the human body contour information and the sensitivity to morphological changes during movement. Key point loss (Ljoi): used to constrain the model to accurately predict the key point information of the human body and the sensitivity to posture changes. The total loss is defined as L=w1×Lfuse+w2×Lsil+w3×Ljoi. The output result is the binary classification result of the total courage level;

[0070] S404: Figure 6 As shown, observe the confusion matrix and analyze the classification accuracy of different dimensions (MT, DT, AT, VS, SB); use the heat map to show the key areas that the model focuses on during the vaulting movement (neck, back, arm and thigh posture areas) to analyze the experimental results and evaluate the performance indicators of the model, including accuracy, precision, recall and F1 score. In summary, the accuracy of the model in predicting the overall courage level is 92%. The reliability of the model is verified through multiple trainings, and the average value of the result accuracy is used to evaluate the performance of the algorithm.

[0071] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest range consistent with the principles and novel features applied for herein.

Claims

1. A sports courage measurement method based on 3D-CNN and Transformer, characterized in that: The following steps are involved: Step 1: SCS-31 was used to measure the level of sports courage. College student volunteers were recruited to complete the questionnaire survey and the SCS-31 was used to assess their sports courage level. Descriptive statistical results were summarized, including the five dimensions of sports courage, namely mastery (MT), determination (DT), assertiveness (AT), venturesome (VS) and self-sacrifice (SB), as well as the mean and standard deviation of the total level and the proportion of females and athletes in the sample. Step 2: Motion data collection, record the volunteer's sports video (taking vaulting as an example), use YOLOV8 to perform human semantic segmentation, obtain key frames, and perform preprocessing such as cropping, cutting out, and aligning the extracted key frames to ensure that the person is in the center of the image. Subsequently, ten key frames representing sub-processes such as run-up, take-off, crossing the horse's back, falling, and landing are randomly selected from the extracted key frames, and the human silhouette is obtained by binarization. YOLOV8-pose is used to extract key point information and generate a key point heat map to reduce the amount of data and improve the accuracy of action recognition. The key point heat map is combined with the human silhouette to obtain two basic modes: the human silhouette and the 3D heat map volume of the key points. Step 3: Establish a data set and process the collected athlete images, including extracting key frames, calculating the center of gravity information and joint angle information, and generating a 2D energy map. The center of gravity information, joint angle information and body swing area in the vaulting horse movement are counted and recorded in a table. Then, the relationship between different dimensions (MT, DT, AT, VS, SB) and the center of gravity information, joint angle information and body swing area in the vaulting horse movement is analyzed. Five of the dimensions are the data measured in step 1. Finally, the center of gravity and joint angle data in the vaulting horse process are combined with the courage information. Step 4: Model construction and training evaluation: S401: A 3D data structure model that integrates 3D-CNN and time-space two-stage Two-Stage-Transformer is constructed, and a routing mechanism is added; S402: Using the cross attention mechanism to fuse the features of the two data modalities of human silhouette image and key points as the input of the model; inputting the input data into the 3D-CNN branch and the Two-Stage-Transformer branch for processing respectively; S403: Calculate the accuracy and loss of using 3D-CNN and Two-Stage-Transformer separately; calculate the accuracy and loss of fusion 3D-CNN and Two-Stage-Transformer. Select a combined loss to train the network, including fusion loss, contour loss, and key point loss; The output result is a binary classification result of high and low total courage level; S404: Observe the confusion matrix and analyze the classification accuracy of different dimensions (MT, DT, AT, VS, SB); use heat maps to display the key areas that the model focuses on during the vaulting movement (neck, back, arm and thigh posture areas) to analyze the experimental results and evaluate the performance indicators of the model, including accuracy, precision, recall and F1 score.

2. The sports courage measurement method based on 3D-CNN and Transformer as claimed in claim 1, characterized in that: Step 2 specifically includes:

1. Apply YOLOV8-pose to each key frame to extract the skeleton key point information of the human body, which is stored in the form of (xk, yk) coordinates, where k represents the kth key point.

2. In order to reduce the interference of small offsets of key points on the network, a Gaussian function is used to generate a key point heat map, and all points around the key points are considered as key points.

3. The parameter σ of the Gaussian function controls the variance of the generated Gaussian image, thereby controlling the smoothness of the heat map.

4. The generated key point heat map presents a Gaussian probability distribution. The grayscale value of the true target point approaches the maximum value 1, while the value around the target point gradually decreases as the distance from the calibrated key point increases.

5. Superimpose all heat maps along the time dimension to obtain the 3D heat map volume of the key point, whose size is K×T×H×W, where K is the number of key points, T is the time dimension, and H and W are the height and width of the frame.

6. The calculation formula of the Gaussian heat map 3. The sports courage measurement method based on 3D-CNN and Transformer as claimed in claim 1, characterized in that: Step 4 specifically includes: fusion loss (Lfuse): used to constrain the consistency and accuracy of the model in the feature fusion process, contour loss (Lsil): used to constrain the model to accurately capture the human body contour information and the sensitivity to morphological changes during movement, key point loss (Ljoi): used to constrain the model to accurately predict the key point information of the human body and the sensitivity to posture changes. The total loss is defined as L = w1×Lfuse+w2×Lsil+w3×Ljoi.