Dynamic gesture recognition method, device and computer-readable storage medium

1. A computer-implemented dynamic gesture recognition method, comprising:

performing detection on each frame of image of a video stream using a preset static gesture detection model to obtain a static gesture in each frame of image of the video stream;

in response to detection of a change of the static gesture from a preset first gesture to a second gesture, suspending the static gesture detection model and activating a preset dynamic gesture detection model; and

performing detection on a plurality of frames of images that are pre-stored in a storage medium using the dynamic gesture detection model to obtain a dynamic gesture recognition result;

wherein suspending the static gesture detection model and activating the preset dynamic gesture detection model comprise:

accumulating a number of frames of images in which the static gesture is the first gesture;

in response to the number being greater than a first threshold, accumulating a number of frames of images in which the static gesture is the second gesture; and

in response to the number of frames of images in which the static gesture is the second gesture being greater than a second threshold, determining that the first gesture has been changed into the second gesture, suspending the static gesture detection model and activating a preset dynamic gesture detection model.

2. The method of claim 1 , further comprising, before performing detection on the plurality of frames of images that are pre-stored in the storage medium using the dynamic gesture detection model to obtain the dynamic gesture recognition result,

storing each frame of image, in which the static gesture is the second gesture before the dynamic gesture detection model is activated, in the storage medium successively; and

storing each frame of image of the video stream after the dynamic gesture detection model is activated until a number of frames of images stored in the storage medium is equal to a third threshold.

3. The method of claim 1 , further comprising, after performing detection on the plurality of frames of images that are pre-stored in the storage medium using the dynamic gesture detection model to obtain the dynamic gesture recognition result,

in response to a number of frames of images counted after the dynamic gesture detection model is activated being equal to a fourth threshold, clearing each frame of images stored in the storage medium.

4. The method of claim 3 , further comprising:

in the initial state, setting a model control flag to a first value that indicates execution of a static gesture detection;

in response to detection of a change of the static gesture from a preset first gesture to a second gesture, setting the model control flag to a second value that indicates execution of a dynamic gesture detection; and

in response to number of frames of images counted after the dynamic gesture detection model is activated being equal to the fourth threshold, setting the model control flag to the first value.

5. The method of claim 1 , wherein the static gesture detection model is a model obtained by using Pelee-SSD to train a preset static gesture training set.

6. The method of claim 1 , wherein the dynamic gesture detection model is a model obtained by using 3D-MobileNetV2 to train a preset dynamic gesture training set.

7. A dynamic gesture recognition device comprising:

one or more processors;

a memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprise:

instructions for performing detection on each frame of image of a video stream using a preset static gesture detection model to obtain a static gesture in each frame of image of the video stream;

instructions for, in response to detection of a change of the static gesture from a preset first gesture to a second gesture, suspending the static gesture detection model and activating a preset dynamic gesture detection model; and

instructions for, performing detection on a plurality of frames of images that are pre-stored in a storage medium using the dynamic gesture detection model to obtain a dynamic gesture recognition result;

wherein the instructions for suspending the static gesture detection model and activating the preset dynamic gesture detection model comprise:

instructions for accumulating a number of frames of images in which the static gesture is the first gesture;

instructions for, in response to the number being greater than a first threshold, accumulating a number of frames of images in which the static gesture is the second gesture; and

instructions for, in response to the number of frames of images in which the static gesture is the second gesture being greater than a second threshold, determining that the first gesture has been changed into the second gesture, suspending the static gesture detection model and activating a preset dynamic gesture detection model.

8. The device of claim 7 , wherein the one or more programs further comprise, before performing detection on the plurality of frames of images that are pre-stored in the storage medium using the dynamic gesture detection model to obtain the dynamic gesture recognition result,

instruction for storing each frame of image, in which the static gesture is the second gesture before the dynamic gesture detection model is activated, in the storage medium successively; and

instruction for storing each frame of image of the video stream after the dynamic gesture detection model is activated until a number of frames of images stored in the storage medium is equal to a third threshold.

9. The device of claim 7 , wherein the one or more programs further comprise, after performing detection on the plurality of frames of images that are pre-stored in the storage medium using the dynamic gesture detection model to obtain the dynamic gesture recognition result,

instruction for, in response to a number of frames of images counted after the dynamic gesture detection model is activated being equal to a fourth threshold, clearing each frame of images stored in the storage medium.

10. The device of claim 9 , wherein the one or more programs further comprise:

instructions for, in the initial state, setting a model control flag to a first value that indicates execution of a static gesture detection;

instructions for, in response to detection of a change of the static gesture from a preset first gesture to a second gesture, setting the model control flag to a second value that indicates execution of a dynamic gesture detection; and

instructions for, in response to number of frames of images counted after the dynamic gesture detection model is activated being equal to the fourth threshold, setting the model control flag to the first value.

11. The device of claim 7 , wherein the static gesture detection model is a model obtained by using Pelee-SSD to train a preset static gesture training set.

12. The device of claim 7 , wherein the dynamic gesture detection model is a model obtained by using 3D-MobileNetV2 to train a preset dynamic gesture training set.

13. A non-transitory computer-readable storage medium storing one or more programs to be executed in a dynamic gesture recognition device, the one or more programs, when being executed by one or more processors of the dynamic gesture recognition device, causing g legged robot to perform processing comprising:

performing detection on each frame of image of a video stream using a preset static gesture detection model to obtain a static gesture in each frame of image of the video stream;

in response to detection of a change of the static gesture from a preset first gesture to a second gesture, suspending the static gesture detection model and activating a preset dynamic gesture detection model; and

performing detection on a plurality of frames of images that are pre-stored in a storage medium using the dynamic gesture detection model to obtain a dynamic gesture recognition result;

wherein suspending the static gesture detection model and activating the preset dynamic gesture detection model comprise:

accumulating a number of frames of images in which the static gesture is the first gesture;

in response to the number being greater than a first threshold, accumulating a number of frames of images in which the static gesture is the second gesture; and

in response to the number of frames of images in which the static gesture is the second gesture being greater than a second threshold, determining that the first gesture has been changed into the second gesture, suspending the static gesture detection model and activating a preset dynamic gesture detection model.

14. The non-transitory computer-readable storage medium of claim 13 , further comprising, before performing detection on the plurality of frames of images that are pre-stored in the storage medium using the dynamic gesture detection model to obtain the dynamic gesture recognition result,

storing each frame of image, in which the static gesture is the second gesture before the dynamic gesture detection model is activated, in the storage medium successively; and

storing each frame of image of the video stream after the dynamic gesture detection model is activated until a number of frames of images stored in the storage medium is equal to a third threshold.

15. The non-transitory computer-readable storage medium of claim 13 , further comprising, after performing detection on the plurality of frames of images that are pre-stored in the storage medium using the dynamic gesture detection model to obtain the dynamic gesture recognition result,

in response to a number of frames of images counted after the dynamic gesture detection model is activated being equal to a fourth threshold, clearing each frame of images stored in the storage medium.

16. The non-transitory computer-readable storage medium of claim 15 , further comprising:

in the initial state, setting a model control flag to a first value that indicates execution of a static gesture detection;

in response to detection of a change of the static gesture from a preset first gesture to a second gesture, setting the model control flag to a second value that indicates execution of a dynamic gesture detection; and

in response to number of frames of images counted after the dynamic gesture detection model is activated being equal to the fourth threshold, setting the model control flag to the first value.

17. The non-transitory computer-readable storage medium of claim 13 , wherein the static gesture detection model is a model obtained by using Pelee-SSD to train a preset static gesture training set.

18. The non-transitory computer-readable storage medium of claim 13 , wherein the dynamic gesture detection model is a model obtained by using 3D-MobileNetV2 to train a preset dynamic gesture training set.

19. The non-transitory computer-readable storage medium of claim 13 , wherein the first threshold is 20.

20. The non-transitory computer-readable storage medium of claim 13 , wherein the second threshold is 10.