Aug 3, 2020
The Shibata Project Group of A03 has published a model for DeepLabCut, which enables the world's first estimation of macaque monkeys' motions in the natural environment from 2D images.
The Shibata Project Group of A3 has been working on “a markerless phenotypic recording and mining system” to discover "individuality" in mice, primates, and humans.
They have recently published a model for DeepLabCut, which can estimate the motion of macaque monkeys in the natural environment from 2D images for the first time in the world, and a preprint on it.
Preprint Figure 1
In this study, they have created a large open dataset of monkey feature labels in natural scenes. This dataset will be useful for training and testing deep learning models for markerless motion capture in macaque monkeys, as well as for the development of algorithms. It is expected to contribute to the establishment of an innovative platform for the analysis of non-human primate animal behavior in many fields such as neuroscience and medicine.
This research was led by the Koseisouhatsu area in collaboration with the National Institutes of Natural Sciences and Kyoto University's Primate Research Institute. The trained model of the DeepLabCut is available in the Model Zoo.