Method, system, and non-transitory computer-readable recording medium for estimating information on golf swing posture

1 . A method performed in a mobile device for estimating information on a golf swing posture, the mobile device comprising one or more processors and the method comprising the steps of:

by the one or more processors, light-weighting an artificial neural network model using depthwise convolution and pointwise convolution to obtain a light-weighted artificial neural network model, wherein a depth of each kernel in the depthwise convolution is reduced to 1 and each of a width and a height of each kernel in the pointwise convolution is reduced to 1;

by the one or more processors, photographing a two-dimensional image of the user's golf swing using a camera;

by the one or more processors, generating a two-dimensional heat map by inputting the two-dimensional image to the light-weighted artificial neural network model, wherein a value of each pixel of the two-dimensional heat map indicates a probability that corresponding pixel corresponds to a joint of the user;

by the one or more processors, estimating a trajectory of a golf ball according to the user's golf swing from the photographed image using the light-weighted artificial neural network model;

by the one or more processors, adjusting a determination criterion for a golf swing of a user on the basis of the two-dimensional heat map;

by the one or more processors, estimating information on a golf swing posture of the user on the basis of the adjusted determination criterion; and

by the one or more processors, pointing out a problem in the user's golf swing posture to the user with reference to the information on the golf swing posture,

wherein in the adjusting step, the determination criterion is adjusted on the basis of the trajectory of the golf ball,

wherein in the adjusting step, a ball flight of the golf ball is determined on the basis of the trajectory of the golf ball, and the determination criterion is adjusted with reference to the determined ball flight, and

wherein in the adjusting step, the determination criterion is more tightly adjusted in response to determining that the user's golf swing posture is more likely to have a problem influencing the ball flight of the golf ball, with reference to the photographed image.

2 . The method of claim 1 , wherein in the adjusting step, the determination criterion is adjusted on the basis of information on the user's actual body, which is estimated from the photographed image using an artificial neural network model.

3 . The method of claim 2 , wherein in the adjusting step, the information on the actual body is estimated with reference to at least one joint of the user detected from the photographed image.

4 . The method of claim 1 , wherein in the adjusting step, the determination criterion is adjusted with reference to the probability.

5 . The method of claim 4 , wherein in the adjusting step, when a probability of the joint being located in a specific position is not greater than a threshold value, the determination criterion is adjusted such that the information on the user's golf swing posture is not estimated.

6 . The method of claim 1 , wherein in the adjusting step, the determination criterion is adjusted with reference to characteristics of a swing posture of a comparison target golfer selected by the user, which are estimated on the basis of information on the comparison target golfer's swing posture.

7 . A non-transitory computer-readable recording medium having stored thereon a computer program for executing the method of claim 1 .

8 . A mobile device for estimating information on a golf swing posture, the mobile device comprising one or more processors configured to:

light-weight an artificial neural network model using depthwise convolution and pointwise convolution to obtain a light-weighted artificial neural network model, wherein a depth of each kernel in the depthwise convolution is reduced to 1 and each of a width and a height of each kernel in the pointwise convolution is reduced to 1;

photograph a two-dimensional image of the user's golf swing using a camera;

generate a two-dimensional heat map by inputting the two-dimensional image to the light-weighted artificial neural network model, wherein a value of each pixel of the two-dimensional heat map indicates a probability that corresponding pixel corresponds to a joint of the user;

estimate a trajectory of a golf ball according to the user's golf swing from the photographed image using the light-weighted artificial neural network model;

adjust a determination criterion for a golf swing of a user on the basis of the two-dimensional heat map;

estimate information on a golf swing posture of the user on the basis of the adjusted determination criterion; and

point out a problem in the user's golf swing posture to the user with reference to the information on the golf swing posture,

wherein the one or more processors are configured to adjust the determination criterion on the basis of the trajectory of the golf ball,

wherein the one or more processors are configured to determine a ball flight of the golf ball on the basis of the trajectory of the golf ball, and adjust the determination criterion with reference to the determined ball flight, and

wherein the one or more processors are configured to more tightly adjust the determination criterion in response to determining that the user's golf swing posture is more likely to have a problem influencing the ball flight of the golf ball, with reference to the photographed image.

9 . The mobile device of claim 8 , wherein the one or more processors are configured to adjust the determination criterion on the basis of information on the user's actual body, which is estimated from the photographed image using an artificial neural network model.

10 . The mobile device of claim 9 , wherein the one or more processors are configured to estimate the information on the actual body with reference to at least one joint of the user detected from the photographed image.

11 . The mobile device of claim 8 , wherein the one or more processors are configured to adjust the determination criterion with reference to the probability.

12 . The mobile device of claim 11 , wherein the one or more processors are configured to, when a probability of the joint being located in a specific position is not greater than a threshold value, adjust the determination criterion such that the information on the user's golf swing posture is not estimated.

13 . The mobile device of claim 8 , wherein the one or more processors are configured to adjust the determination criterion with reference to characteristics of a swing posture of a comparison target golfer selected by the user, which are estimated on the basis of information on the comparison target golfer's swing posture.