Method and device for recommending golf-related contents, and non-transitory computer-readable recording medium
A method for recommending golf-related contents is provided. The method includes detecting at least one body part of a user from a captured image by using an artificial neural network model, when the captured image related to the golf swing of the user is acquired; estimating at least one problem about the golf swing of the user by referring to the position of the detected at least one body part; and deriving, in reference to information related to the problem associated with the golf-related contents, at least one piece of candidate content associated with the at least one problem from among golf-related contents.
1. A method for recommending golf instructional videos, the method comprising the steps of:
in response to acquiring a photographed image of a golf swing of a user, detecting at least one body part of the user from the photographed image using an artificial neural network model;
estimating at least one problem of the golf swing of the user based on a position of the at least one detected body part;
recognizing words in audio data associated with partial motions of golf swings in golf instructional videos by using a recurrent neural network model;
estimating one or more problems associated with a golf swing in each of the golf instructional videos based on the recognized words;
determining candidate golf instructional videos among the golf instructional videos by comparing the estimated at least one problem of the golf swing of the user and the estimated one or more problems associated with the golf swing in each of the golf instructional videos; and
providing one or more of the candidate golf instructional videos to a device of the user.
2. The method of claim 1 , wherein the artificial neural network model is light-weighted using depthwise convolution and pointwise convolution.
3. The method of claim 1 , wherein in the estimating step, information on a golf club is estimated with reference to the position of the at least one detected body part, and the at least one problem is estimated with further reference to the estimated information on the golf club.
4. The method of claim 1 , wherein at least one of the at least one problem and the one or more problems associated with a golf swing in each of the golf instructional videos is estimated separately for each partial motion constituting the golf swing.
5. The method of claim 1 , further comprising the step of:
before the photographed image is acquired, establishing parameters of a photographing module such that motion blur is reduced.
6. The method of claim 1 , wherein the one or more problems associated with the golf swing in each of the golf instructional videos is estimated by analyzing problem areas derived from the golf instructional videos.
7. The method of claim 6 , wherein the problem areas are derived with reference to at least one of video data and audio data of the golf instructional videos.
8. The method of claim 1 , wherein the one or more problems associated with the golf swing in each of the golf instructional videos is further estimated with reference to at least one of metadata, video data, and audio data of the golf instructional videos.
9. A non-transitory computer-readable recording medium having stored thereon a computer program for executing the method of claim 1 .
10. The method of claim 1 , further comprising:
calculating a matching score for each of the candidate golf instructional videos; and
providing the one or more of the candidate golf instructional videos to the device of the user based on the calculated matching scores.
11. The method of claim 10 ,
wherein the matching score for each of the candidate golf instructional video is calculated based on a number of problems matching between the at least one problem of the golf swing of the user and the one or more problems associated with a golf swing in each of the golf instructional videos.
12. The method of claim 1 , wherein:
the one or more problems associated with the golf swing in each of the golf instructional videos are estimated further based on metadata of the golf instructional videos.
13. The method of claim 1 , wherein each of the partial motions is one of an address, a takeaway, a back swing, a top-of-swing, a down swing, an impact, a follow-through, and a finish.
14. A system for recommending golf instructional videos, the system comprising:
a body part detection unit configured to, in response to acquiring a photographed image of a golf swing of a user, detect at least one body part of the user from the photographed image using an artificial neural network model;
a problem estimation unit configured to estimate at least one problem of the golf swing of the user based on a position of the at least one detected body part;
a content management unit configured to:
recognize words in audio data associated with partial motions of golf swings in golf instructional videos by using a recurrent neural network model;
estimate one or more problems associated with a golf swing in each of the golf instructional videos based on the recognized words; and
determine candidate golf instructional videos among the golf instructional videos by comparing the estimated at least one problem of the golf swing of the user and the estimated one or more problems associated with the golf swing in each of the golf instructional videos; and
a communication unit configured to transmit one or more of the candidate golf instructional videos to a device of the user.
15. The system of claim 14 , wherein the artificial neural network model is light-weighted using depthwise convolution and pointwise convolution.
16. The system of claim 14 , wherein the problem estimation unit is configured to estimate information on a golf club with reference to the position of the at least one detected body part, and estimate the at least one problem with further reference to the estimated information on the golf club.
17. The system of claim 14 , wherein at least one of the at least one problem and the one or more problems associated with a golf swing in each of the golf instructional videos is estimated separately for each partial motion constituting the golf swing.
18. The system of claim 14 , wherein the body part detection unit is configured to, before the photographed image is acquired, establish parameters of a photographing module such that motion blur is reduced.
19. The system of claim 14 , wherein the one or more problems associated with the golf swing in each of the golf instructional videos is estimated by analyzing problem areas derived from the golf instructional videos.
20. The system of claim 14 , wherein the one or more problems associated with the golf swing in each of the golf instructional videos is estimated with reference to at least one of metadata, video data, and audio data of the golf instructional videos.
21. The system of claim 19 , wherein the problem areas are derived with reference to at least one of video data and audio data of the golf instructional videos.