Fig 1.
Outline of the fatiguing protocol.
The protocol was repeated until the player reached volitional exhaustion. The 20-s rest was decreased by 5 s after the fifth and tenth cycles. The dashed box shows the section of each cycle that was captured by the markerless system.
Fig 2.
Diagram of the experimental set-up on the tennis court.
Black crosses show approximate camera positions; the yellow cross shows the serving position; yellow squares show the ball target zones (player moves to the left after serving); red arrows show directions of the sprint (starting from the right target zone and finishing at the net).
Fig 3.
Pose estimation keypoint locations.
Based on the COCO-Wholebody dataset [38].
Table 1.
Descriptive results of the fatiguing protocol, giving the number of cycles completed, peak velocity during the baseline sprint test, reduction in peak sprint velocity across the protocol and average and maximum heart rates across the whole protocol.
Table 2.
Mean (± SD) of the positive mechanical work done during each cycle of the fatiguing protocol and the cumulative positive mechanical work across the entire protocol.
Fig 4.
Peak sprint velocity of each cycle against cumulative positive mechanical work for each player.
Each data point corresponds to a completed cycle of the protocol. Pearson’s r coefficients given to the side of each plot. (A) Total mechanical work. (B) External work only. (C) Pelvis approximation. (D) Bounding box approximation.
Fig 5.
Repeated measures Bland-Altman of positive external work, comparing the centre of mass (CoM) proxy methods (pelvis and bounding box) against the CoM work.
Each player is represented by a different colour. Pelvis method is represented by circles and dotted lines. Bounding box method is represented by crosses and dashed lines. (Note: For clarity, the scale of the y-axis is not uniform across the two sections of the figure.).
Fig 6.
Example of energies during one 30-s trial, for the centre of mass and proxy methods.