The Blackbird unmanned aerial vehicle (UAV) dataset is a large-scale,\naggressive indoor flight dataset collected using a custom-built quadrotor\nplatform for use in evaluation of agile perception.Inspired by the potential of\nfuture high-speed fully-autonomous drone racing, the Blackbird dataset contains\nover 10 hours of flight data from 168 flights over 17 flight trajectories and 5\nenvironments at velocities up to $7.0ms^-1$. Each flight includes sensor data\nfrom 120Hz stereo and downward-facing photorealistic virtual cameras, 100Hz\nIMU, $\\sim190Hz$ motor speed sensors, and 360Hz millimeter-accurate motion\ncapture ground truth. Camera images for each flight were photorealistically\nrendered using FlightGoggles across a variety of environments to facilitate\neasy experimentation of high performance perception algorithms. The dataset is\navailable for download at http://blackbird-dataset.mit.edu/\n