DeepLabStream enables closed-loop behavioral experiments using deep learning-based markerless, real-time posture detection

• DOI: 10.1038/s42003-021-01654-9
• Publication Date: 2021-01-29T11:19:10Z
• Authors: Jens F. Schweihoff; Matvey Loshakov; Irina Pavlova; Laura Kück; Laura A. Ewell; Martin K. Schwarz
• Keywords: Millisecond; Offline learning

Abstract In general, animal behavior can be described as the neuronal-driven sequence of reoccurring postures through time. Most available current technologies focus on offline pose estimation with high spatiotemporal resolution. However, to correlate neuronal activity it is often necessary detect and react online behavioral expressions. Here we present DeepLabStream, a versatile closed-loop tool providing real-time deliver posture dependent stimulations. DeepLabStream has temporal resolution in millisecond range, utilize different input, well output devices tailored multiple experimental designs. We employ semi-autonomously run second-order olfactory conditioning task freely moving mice optogenetically label ensembles active during specific head directions.