Drone Videos

### Overview
The dataset is a number of videos recorded with the DJI Mavic Pro in various environments with and without people present.

### Ideas
The idea for the dataset is to apply various analysis algorithms like object detection, motion tracking, SLAM, semantic segmentation and others to real data. The dataset has no ground-truth labels, but is fairly representtive of what real drone footage looks like and hasn't be significantly post-processed (DJI has quite a bit of post-processing built in to change colors)

### 概述
本数据集由多段使用 DJI Mavic Pro 在多种环境中拍摄的视频组成，其中包括有人和无人在场的情况。

### 研究思路
本数据集的设计理念在于，通过将各类分析算法，如目标检测、运动追踪、同步定位与地图构建（SLAM）、语义分割等应用于实际数据，以实现对真实无人机视频数据的分析。该数据集缺乏地面真实标签，但足以体现真实无人机视频的典型特征，且未经显著后期处理（DJI内置了相当数量的后期处理功能，用于调整色彩）。