JP2022124826A - Evaluation device, method and program - Google Patents

Abstract

To provide an evaluation device, a method and a program, capable of improving evaluation accuracy of a faint action.SOLUTION: An evaluation device 10 includes an acquisition unit 11 and an evaluation unit 12. The acquisition unit 11 acquires an imaged faint action of a player. The evaluation unit 12 evaluates the faint action based on a model learned by machine learning, stored beforehand. In this case, the evaluation unit 12 evaluates the faint action based on whether or not the action is different from a prediction operation predicted by the learned model.SELECTED DRAWING: Figure 1

Description

The present invention relates to an evaluation device, method and program.

A sports player obtains information such as a feint method from a reference book or other people's play, and tries it out in practice. A player with good taste can embody the play just by watching it, but many players cannot embody it. Also, when a player receives advice from a coach, it is difficult for the player to know how his play has improved.
Patent Literature 1 discloses a technique for objectively evaluating a motion such as a player's play or performance during exercise using a machine-learned model.

JP 2018-68516 A

In Patent Literature 1, the completeness of a player's action is evaluated according to a learning model in which actions (plays, performances) and scores are linked. Here, among plays, a feint movement is a movement that confuses an opponent, and its perfection is highly evaluated when the opponent's prediction is betrayed. However, Patent Document 1 does not disclose this content. Therefore, in Patent Literature 1, there is a problem that there is room for improvement in the evaluation accuracy of the feint motion.

An object of the present disclosure is to provide an evaluation apparatus, method, and program capable of improving the evaluation accuracy of feint motions by solving such problems.

The evaluation device of the present disclosure includes an acquisition unit that acquires an imaged feint motion of a player, an evaluation unit that evaluates the feint motion based on a pre-stored learned model by machine learning, and the evaluation unit: The feint action is evaluated based on whether it differs from the predicted action predicted by the trained model.

A method of the present disclosure includes acquiring an imaged feint motion of a player, and evaluating the feint motion based on a pre-stored learned model by machine learning, and evaluating the feint motion. Doing includes evaluating the feint action based on whether it differs from a predicted action predicted by the trained model.

A program according to the present disclosure causes a computer to perform a process of acquiring an imaged feint motion of a player, and a process of evaluating the feint motion based on a pre-stored model trained by machine learning. The process of evaluating the motion includes the process of evaluating the feint motion based on whether it differs from the predicted motion predicted by the trained model.

According to the present disclosure, it is possible to provide an evaluation device, method, and program capable of improving the evaluation accuracy of feint motions.

1 is a block diagram showing the configuration of an evaluation device according to a first embodiment; FIG. It is a schematic diagram which shows the outline of a structure of the evaluation apparatus concerning 2nd Embodiment. It is a block diagram showing the details of the configuration of the evaluation device according to the second embodiment. 9 is a flowchart showing evaluation operations of the evaluation device according to the second embodiment; 9 is a flow chart showing a scoring method of the evaluation device according to the second embodiment; It is a schematic diagram which shows an example of the scoring method of the evaluation apparatus concerning 2nd Embodiment. 9 is a flowchart showing learning operation of the evaluation device according to the second embodiment; It is a block diagram which shows the structure of the computer of this embodiment.

Hereinafter, specific embodiments to which the present disclosure is applied will be described in detail with reference to the drawings. In each drawing, the same elements are denoted by the same reference numerals, and redundant description will be omitted as necessary for clarity of description.

(First embodiment)
First, the configuration of the evaluation device 10 according to the first embodiment will be described with reference to FIG.
The evaluation device 10 includes an acquisition unit 11 and an evaluation unit 12 . The acquisition unit 11 acquires the captured feint motion of the player. The evaluation unit 12 evaluates the player's feint motion based on a model that has been learned by machine learning and is stored in advance.

Next, operation of the evaluation device 10 according to the first embodiment will be described.
First, the acquisition unit 11 acquires the captured feint motion of the player. Next, the evaluation unit 12 evaluates the player's feint motion based on a machine-learned model that is stored in advance. Specifically, the evaluation unit 12 evaluates the feint motion based on whether it differs from the predicted motion predicted by the trained model.

As described above, the evaluation apparatus 10 according to the first embodiment can evaluate a feint action based on whether it is different from the predicted action by machine learning (whether the prediction was betrayed). Therefore, the evaluation device 10 can improve the evaluation accuracy of the feint motion.

(Second embodiment)
First, the configuration of the evaluation device 20 according to the second embodiment will be described with reference to FIGS. 2 and 3. FIG. As shown in FIGS. 2 and 3, the evaluation device 20 is a specific embodiment of the evaluation device 10 according to the first embodiment, and includes an imaging unit 21, an acquisition unit 22, an evaluation unit 23, an output unit 24, a learning A unit 25 and a storage unit 26 are provided. The acquisition unit 22 and the evaluation unit 23 correspond to the acquisition unit 11 and the evaluation unit 12 according to the first embodiment, respectively.

The image capturing unit 21 has an image capturing function such as a camera, and captures an image (a series of multiple images) or a moving image of the feint action of the player to be evaluated. For example, the image capturing unit 21 captures an image of the action of a single player during practice from the front by the operation of the user of the evaluation device 20 . Then, the imaging unit 21 supplies the captured image to the acquisition unit 22 .

The acquisition unit 22 acquires the captured image from the imaging unit 21 . Also, the captured image stored in the storage unit 26 may be obtained. Then, the acquisition unit 22 supplies the acquired captured image to the evaluation unit 23 when the evaluation unit 23 executes the evaluation operation for evaluating the player's feint operation. Further, the acquisition unit 22 supplies the acquired captured image to the learning unit 25 when the learning unit 25 executes a learning operation for learning a feint operation of the player.

The evaluation unit 23 executes evaluation operations in the evaluation device 20 . A captured image is acquired from the acquisition unit 22 . The evaluation unit 23 evaluates the player's feint motion by, for example, scoring it based on the captured image and the learned model stored in advance in the storage unit 26 .

Here, the evaluation unit 23 evaluates the feint motion based on whether it differs from the predicted motion predicted by the trained model. Specifically, the evaluation unit 23 extracts the first action in the feint motion. An action is a motion performed by the player in a series of feint motions, such as acceleration to the left or right, tilting the state left or right, stepping over the ball, changing the line of sight, or the like. Then, the evaluation unit 23 evaluates the case where the second action to be executed next to the first action is different from the predicted action predicted by the trained model, and to be higher than the evaluation of Then, the evaluation unit 23 totalizes the scores of the series of feint actions and calculates the score.
In addition, the evaluation unit 23 may use a learned model to evaluate the feint motion based on the degree of similarity between the feint motion and the feint motion of a player whose skill is equal to or higher than a predetermined level.

The output unit 24 outputs the score calculated by the evaluation unit 23 to a display or the like. Further, the output unit 24 may output the calculated score to an external device such as a mobile terminal connected or communicating with the evaluation device 20 . By doing so, the user of the evaluation device 20 can quantitatively evaluate the player's feint motion.

The learning unit 25 executes evaluation operations in the evaluation device 20 . The learning unit 25 acquires the captured image from the acquisition unit 22 . Then, the learning unit 25 generates a learning model for scoring the feint motion. A learning model for which learning has been completed is hereinafter referred to as a learned model.
The storage unit 26 stores the learning model generated by the learning unit 25 and the learned model. Note that the storage unit 26 may store images or moving images of the player captured by the imaging unit 21 .

Subsequently, the evaluation operation of the evaluation device 20 according to the second embodiment will be described with reference to FIG.
As shown in FIG. 4, first, the imaging unit 21 has an imaging function such as a camera, and captures an image (a series of multiple images) or a moving image of the player's feint action. Then, the acquiring unit 22 acquires the captured image from the imaging unit 21 (step S101). Furthermore, the acquisition unit 22 supplies the acquired captured image to the evaluation unit 23 .

Next, the evaluation unit 23 uses the received captured image and the learned model stored in the storage unit 26 to score the player's feint motion (step S102).
For example, in step S102, as shown in FIG. 5, the evaluation unit 23 evaluates the feint action based on whether it differs from the predicted action predicted by the learned model. Specifically, the evaluation unit 23 detects an arbitrary first action in the feint motion included in the received captured image (step S201). Next, the evaluation unit 23 extracts a second action following the detected first action (step S202).

Next, the evaluation unit 23 scores the player's feint motion based on whether or not the actual next action betrays the second action (step S203). Specifically, the evaluation unit 23 gives a higher evaluation when the second action is different from the predicted action predicted by the trained model than when the second action is the same as the predicted action. For example, as shown in FIG. 6, the trained model predicts 10% of the time that the player will accelerate to the right on the second action, or 90% of the time that the player will accelerate to the left on the second action. . In that case, the evaluation unit 23 gives 9 points when the player actually accelerates to the right in the second action. On the other hand, the evaluation unit 23 gives 1 point when the player actually accelerates to the left in the second action. Note that the score value is not limited to the value described above, and when the player actually accelerates to the right in the second action, 9 points may be calculated as 90 points.

The evaluation unit 23 may repeat the operations of steps S201 to S203 in the same feint operation. In that case, the evaluation unit 23 may calculate the score by averaging or adding the total score of the series of actions in the feint motion.

Finally, the output unit 24 outputs the score calculated by the evaluation unit to a display or the like (step S103). For example, the evaluation unit 23 can score feint actions for each individual, such as 7.5 points for Mr. A, 8.5 points for Mr. B, and 9.0 points for Mr. C. You can output the score calculated in .

Next, the learning operation of the evaluation device 20 according to the second embodiment will be described with reference to FIG.
As shown in FIG. 7, first, the acquisition unit 22 acquires teacher data (step S301). The user of the evaluation device 20 uses, for example, images and moving images of feint movements performed by the player as training data. Next, the learning unit 25 learns the feint motion from the teacher data acquired by the acquisition unit 22, and the above-described evaluation unit 23 learns the feint motion of the player and generates a learning model for scoring the feint motion. (step S302). Next, the learning unit 25 stores the generated learning model in the storage unit 26 (step S303). Here, the evaluation device 20 may repeatedly acquire teacher data in step S201 to grow the learning model. The finally constructed learning model is called a trained model.

As with the evaluation apparatus 10 according to the first embodiment, the evaluation apparatus 20 according to the second embodiment can evaluate the feint movement based on whether it differs from the predicted movement based on machine learning. Therefore, the evaluation device 20 can improve the evaluation accuracy of the feint motion.
In addition, the evaluation device 20 feeds back the scored feint actions of the player. By doing so, the user of the evaluation device 20 can compare the quality of feint motions between players. Also, the player can objectively evaluate his/her own assignment.

It should be noted that the present invention is not limited to the above embodiments, and can be modified as appropriate without departing from the scope of the invention.

(Modification 1)
For example, in step S102, the evaluation unit 23 may use a learned model to score the feint motion based on the degree of similarity between the feint motion and the feint motion of a player whose skill is equal to or higher than a predetermined level.

Specifically, in step S301, the learning unit 25 of Modification 1 generates a trained model using, for example, images and videos of feint actions performed by a player who is good at feints as teacher data. Then, the evaluation unit 23 calculates the degree of similarity between the feint motion of the player and the feint motion of the player whose skill is equal to or higher than a predetermined level, using a predetermined index from the feint motion included in the received captured image. Predetermined indicators are, for example, the overall movement of the player, the line of sight, the direction of the body, the movement of the hand, the movement of the foot, how the movement is slowed down, the position of the ball, and the like. Then, the evaluation unit 23 scores the player's feint motion based on the degree of similarity, such as 9.0 points if the degree of similarity is 90%, and reflects it in the scores calculated in steps S201 to S203. good too. Alternatively, the evaluation unit 23 may perform the evaluation based on the degree of similarity described above without performing the operations from step S201 to step S203.

<Hardware configuration>
Next, a hardware configuration example of the computer 1000 associated with the evaluation device 10 and the evaluation device 20 will be described with reference to FIG. A computer 1000 in FIG. 8 has a processor 1001 and a memory 1002 . The processor 1001 may be, for example, a microprocessor, an MPU (Micro Processing Unit), or a CPU (Central Processing Unit). Processor 1001 may include multiple processors. Memory 1002 is comprised of a combination of volatile and non-volatile memory. Memory 1002 may include storage remotely located from processor 1001 . In this case, processor 1001 may access memory 1002 via an I/O interface (not shown).

Further, each device in the above-described embodiments is configured by hardware or software, or both, and may be configured by one piece of hardware or software, or may be configured by multiple pieces of hardware or software. The functions (processing) of each device in the above-described embodiments may be implemented by a computer. For example, a program for performing the method in the embodiment may be stored in the memory 1002 and each function may be realized by executing the program stored in the memory 1002 with the processor 1001 .

These programs can be stored and delivered to computers using various types of non-transitory computer readable media. Non-transitory computer-readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (eg, flexible discs, magnetic tapes, hard disk drives), magneto-optical recording media (eg, magneto-optical discs), CD-ROMs (Read Only Memory), CD-Rs, CD-R/W, semiconductor memory (eg, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (random access memory)). The program may also be delivered to the computer on various types of transitory computer readable medium. Examples of transitory computer-readable media include electrical signals, optical signals, and electromagnetic waves. Transitory computer-readable media can deliver the program to the computer via wired channels, such as wires and optical fibers, or wireless channels.

10 evaluation device 11 acquisition unit 12 evaluation unit 20 evaluation device 21 imaging unit 22 acquisition unit 23 evaluation unit 24 output unit 25 learning unit 26 storage unit

Claims (8)

an acquisition unit that acquires an imaged feint movement of the player;
an evaluation unit that evaluates the feint motion based on a pre-stored model learned by machine learning;
The evaluation device, wherein the evaluation unit evaluates the feint action based on whether or not it differs from a predicted action predicted by the learned model. The evaluation device according to claim 1, wherein the evaluation unit evaluates the feint action different from the predicted action predicted by the learned model higher than the feint action similar to the predicted action. The evaluation unit extracts a first action in the feint action, and evaluates when a second action to be executed next to the first action is different from the predicted action predicted by the learned model, 3. The evaluation device according to claim 1, wherein the evaluation is made higher than when the second action is similar to the predicted action. 4. The evaluation unit further evaluates the feint motion based on a degree of similarity between the feint motion and a feint motion of a player whose skill is equal to or higher than a predetermined level, using the learned model. The evaluation device according to item 1. The evaluation device according to any one of claims 1 to 4, wherein the evaluation unit evaluates the feint action based on a score. The evaluation device according to any one of claims 1 to 5, further comprising an output unit that outputs the evaluation result for each player. Acquiring an imaged feint movement of the player;
Evaluating the feint action based on a pre-stored model trained by machine learning,
Evaluating the feint action includes evaluating the feint action based on whether it differs from a predicted action predicted by the trained model. a process of acquiring an imaged feint movement of the player;
causing a computer to execute a process of evaluating the feint motion based on a pre-stored model trained by machine learning;
A program, wherein the process of evaluating the feint action includes a process of evaluating the feint action based on whether or not it differs from a predicted action predicted by the learned model.

Images