Drone-Action: An Outdoor Recorded Drone Video Dataset for Action Recognition

摘要

Aerial human action recognition is an emerging topic in drone applications. Commercialdrone platforms capable of detecting basic human actions such as hand gestures have been developed.However, a limited number of aerial video datasets are available to support increased research intoaerial human action analysis. Most of the datasets are confined to indoor scenes or object tracking andmany outdoor datasets do not have sufficient human body details to apply state-of-the-art machinelearning techniques. To fill this gap and enable research in wider application areas, we present anaction recognition dataset recorded in an outdoor setting. A free flying drone was used to record 13dynamic human actions. The dataset contains 240 high-definition video clips consisting of 66,919frames. All of the videos were recorded from low-altitude and at low speed to capture the maximumhuman pose details with relatively high resolution. This dataset should be useful to many researchareas, including action recognition, surveillance, situational awareness, and gait analysis. To testthe dataset, we evaluated the dataset with a pose-based convolutional neural network (P-CNN)and high-level pose feature (HLPF) descriptors. The overall baseline action recognition accuracycalculated using P-CNN was 75.92%.