Markerless 2D kinematic analysis of underwater running: A deep learning approach

Abstract

Kinematic analysis is often performed with a camera system combined with reflective markers placed over bony landmarks. This method is restrictive (and often expensive), and limits the ability to perform analyses outside of the lab. In the present study, we used a markerless deep learningbased method to perform 2D kinematic analysis of deepwater running, a task that poses several challenges to image processing methods. A single GoPro camera recorded sagittal plane lower limb motion. A deep neural network was trained using data from 17 individuals, and then used to predict the locations of markers that approximated joint centres. We found that 300-400 labelled images were sufficient to train the network to be able to position joint markers with an accuracy similar to that of a human labeler (mean difference <3 pixels, around 1cm). This level of accuracy is sufficient for many 2D applications, such as sports biomechanics, coaching/training, and rehabilitation. The method was sensitive enough to differentiate between closely-spaced running cadences (45-85 strides per minute in increments of 5). We also found high test-retest reliability of mean stride data, with between-session correlation coefficients of 0.90-0.97. Our approach represents a low-cost, adaptable solution for kinematic analysis, and could easily be modified for use in other movements and settings. Using additional cameras, this approach could also be used to perform 3D analyses.The method presented here may have broad applications in different fields, for example by enabling markerless motion analysis to be performed during rehabilitation, training or even competition environments.

Moving outside the lab: markerless motion capture accurately quantifies sagittal plane kinematics during the vertical jump

“…Previous studies have described the technical aspects of model training in detail (Cronin et al, 2019; Mathis and Mathis, 2020). Therefore, we will focus on how we trained the model in practice while providing a detailed human biomechanics-focused user guide for reference in appendices.…”

“…Rather than tracking a predetermined list of external features or landmarks, it can track any feature of interest. It has been used in disparate applications ranging from tracking the muscle-tendon junction during muscle contractions imaged using ultrasound (Krupenevich et al, 2020) to tracking human movement outside of the laboratory in challenging conditions like underwater running (Cronin et al, 2019). However, this tool has yet to be widely used in the field of biomechanics.…”

Pose2Sim: An End-to-End Workflow for 3D Markerless Sports Kinematics—Part 1: Robustness

“…They showed that OpenPose is robust to challenging clothing such as large Indian pants, as well as to extreme lightning conditions. Other sports activities have been investigated, such as lower body kinematics of vertical jumps [ 28 ] or underwater running [ 29 ]. Both works trained their own model with DeepLabCut.…”

Human Movement Science in The Wild: Can Current Deep-Learning Based Pose Estimation Free Us from The Lab?

“…As such, we would expect to see improved results for DeepLabCut if additional data and training time is leveraged on top of the DeepLabCut pre-trained human pose model that was evaluated in this study. Indeed, previous work utilising re-trained DeepLabCut has demonstrated promising 2D sagittal plane results during underwater running with mean differences of approximately 10 mm 45 .…”