Comparison of the Performance of DeepLabCut models trained with Different Number of Participant Data

DeepLabCut (DLC) (Mathis et al., 2018) is a convolutional neural network technique for markerless pose estimation. DLC has been shown to achieve similar accuracy to traditional motion capture systems in specific tasks (Kosourikhina et al., 2022). However, the DLC network model requires participant data, often from multiple individuals (Mathis etal., 2020), making the process time-consuming. Therefore, it is crucial to determine the minimum number and fraction of participants necessary for model training to enhance the feasibility of DLC for a wide range of users. In this study, we investigated the performance of DLC’s network models by comparing their performance when trained with one, four, and eight participants.     «

DeepLabCut (DLC) (Mathis et al., 2018) is a convolutional neural network technique for markerless pose estimation. DLC has been shown to achieve similar accuracy to traditional motion capture systems in specific tasks (Kosourikhina et al., 2022). However, the DLC network model requires participant data, often from multiple individuals (Mathis etal., 2020), making the process time-consuming. Therefore, it is crucial to determine the minimum number and fraction of participants necessary for mod...     »