Estimating spin rate and axis of a ball using deep learning

Patent metadata

Field	Value
Publication number	US-12582870-B2
Application number	US-202418668314-A
Country	US
Kind code	B2
Filing date	May 20, 2024
Priority date	May 20, 2024
Publication date	Mar 24, 2026
Grant date	Mar 24, 2026

Abstract

Embodiments are disclosed for determining a spin rate and axis of a ball using deep learning. In some embodiments, a method comprises: training a deep learning network on training images of spinning balls, each spinning ball having at least one feature point in a time series that forms a two-dimensional (2D) ellipse image in a 2D plane; capturing, with at least one camera, a series of images of a ball; predicting, with the trained deep learning network, spin measurements associated with the ball based on the series of images; determining a spin rate of the ball based on the spin measurements; determining coefficients of a 2D ellipse model based on the spin measurements and the spin rate; and determining, with the at least one processor, a spin axis of the ball in 3D space based on the 2D ellipse model and the spin rate.

First claim

The invention claimed is:

1 . A method comprising:
at a first time, training a deep learning network on a series of training images of spinning balls, each spinning ball having at least one feature point in a time series that forms a three-dimensional (3D) circle in 3D space and a corresponding two-dimensional (2D) ellipse image in a 2D plane; at a second time after the first time, capturing, with at least one camera, a series of images of a ball; predicting, with the trained deep learning network, spin measurements associated with the ball based on the series of images; determining, with at least one processor, a spin rate of the ball based on the spin measurements; determining, with the at least one processor, coefficients of a 2D ellipse model based on the spin measurements and the spin rate; and determining, with the at least one processor, a spin axis of the ball in 3D space based on the 2D ellipse model and the spin rate.

2 . The method of claim 1 , wherein the spin rate is computed from a norm of the spin measurements.

3 . The method of claim 1 , wherein the trained deep learning network generates a spin confidence score associated with the spin measurements.

4 . The method of claim 3 , wherein the deep learning network comprises n units, each unit including a convolutional filter followed by an activation function, and a fully connected layer for outputting the spin measurements and the spin confidence score.

5 . The method of claim 4 , wherein the activation function is a rectified linear unit (ReLU).

6 . The method of claim 1 , wherein the training images include different types of balls having different spin rates.

7 . The method of claim 6 , wherein the training images are augmented by at least one of color, contrast, brightness, sharpness, shift or noise level, non-uniform lighting or complex backgrounds.

8 . The method of claim 6 , wherein at least one of the training images is randomly augmented.

9 . The method of claim 1 , wherein the at least one feature point is a dot, scratch, dimple or seam on the ball.

10 . The method of claim 1 , further comprising:
creating a 2D ellipse image from the series of images; and inputting the 2D ellipse image into the trained deep learning network.

11 . A system comprising:
at least one camera that captures a series of images of a ball; at least one processor configured to:
predict, with a deep learning network, spin measurements associated with the ball based on the series of images, where the deep learning network is trained on a series of training images of spinning balls, each spinning ball having at least one feature point in a time series that forms a three-dimensional (3D) circle in 3D space and a corresponding two-dimensional (2D) ellipse image in a 2D plane;
determine a spin rate of the ball based on the spin measurements;
determine coefficients of a two-dimensional (2D) ellipse model based on the spin measurements and the spin rate; and
determine a spin axis of the ball in three-dimensional (3D) space based on the ellipse model and spin rate.

12 . The system of claim 11 , wherein the spin rate is computed from a norm of the spin measurements.

13 . The system of claim 11 , wherein the trained deep learning network generates a spin confidence score associated with the spin measurements.

14 . The system of claim 13 , wherein the deep learning network comprises n units, each unit including a convolutional filter followed by an activation function, and a fully connected layer for outputting the spin measurements and the spin confidence score.

15 . The system of claim 14 , wherein the activation function is a rectified linear unit (ReLU).

16 . The system of claim 11 , wherein the training images include different types of balls having different spin rates.

17 . The system of claim 16 , wherein the training images are augmented by at least one of color, contrast, brightness, sharpness, shift or noise level, non-uniform lighting or complex backgrounds.

18 . The system of claim 17 , wherein at least one of the training images is randomly augmented.

19 . The system of claim 16 , wherein the at least one feature point includes at least one of a dot, scratch, dimple or seam on the ball.

20 . The system of claim 11 , further comprising:
creating a 2D ellipse image from the series of images; and inputting the 2D ellipse image into the trained deep learning network.

Assignees

• Rapsodo Pte Ltd

Inventors

• Xu Dejiang
• Keat Kelvin Yeo Soon
• Okur Batuhan

Classifications

• A63B69/3623

  for driving (A63B69/3608, A63B69/3658, A63B69/3661, A63B69/3667 and A63B69/3691 take precedence) · CPC title

• A63B24/0006

  Computerised comparison for qualitative assessment of motion sequences or the course of a movement · CPC title

• A63B69/3658

  Means associated with the ball for indicating or measuring, e.g. speed, direction · CPC title

• A63B2220/35

  Spin · CPC title

• G06T2207/20081

  Training; Learning · CPC title