Skip to main content
Dryad

Gerbil family video and pose tracking files with metadata

Mitelut, Catalin1; Sanes, Dan2

Research facility: New York University
Published Sep 05, 2025 on Dryad. https://doi.org/10.5061/dryad.pc866t1vk

Data files

Sep 05, 2025 version files 246.55 GB

Click names to download individual files Select up to 11 GB of files for zip download

Abstract

Understanding animal behavior requires the capability of monitoring and quantifying behaviors across many time points and increasingly ethological behaviors.  In the past several years, machine vision researchers have made significant advances in building animal tracking software relying on the development of convolutional neural networks (CNNs) and deep learning. However, the lack of longitudinal, curated video datasets for machine learning and behavioral science remains a key bottleneck in advancing both fields. Here, we present 3 curated multi-animal, multi-day rodent behavior datasets containing the behavior of gerbil families (2 adults and 4 pups) during critical periods of development (P15 to P30). Each cohort contains videos acquired at 24 fps for 15 days continuously, yielding approximately 30 million frames per cohort. We provide ground truth annotations for all animals in 1000 to 6000 frames across the cohorts using a standardized Ultralytics pose data format.  We additionally provide predictions from using Sleap (Pereira et al, 2022) machine vision tracking as a dataset to be used by behavioral scientists for analysis and machine vision researchers as a benchmark.