开放非正式数据集影响自动驾驶

本数据集由光州科学技术研究院的GIST自动驾驶实验室创建，旨在为自动驾驶车辆的对象识别技术开发提供道路上的物体和非结构化动态数据。数据集包含300,000张图像，这些图像是从总计1,600小时的驾驶图像中提取的关键帧，涵盖了多种环境条件，如四季变化、不同天气和时间。数据收集过程中采用了两种方法：车辆前部安装摄像头和使用三脚架固定摄像头。数据集的创建过程包括原始数据处理、数据信息展示方法、注释数据展示方法等。该数据集主要应用于自动驾驶领域，解决对象识别和环境感知的问题。

This dataset was created by the GIST Autonomous Driving Laboratory at the Gwangju Institute of Science and Technology. It is designed to provide on-road objects and unstructured dynamic data for the development of object recognition technologies for autonomous vehicles. The dataset contains 300,000 images, which are key frames extracted from a total of 1,600 hours of driving footage, covering diverse environmental conditions including seasonal changes, various weather scenarios, and different times of day. Two methods were adopted for data collection: mounting cameras on the front of the vehicle and using tripod-fixed cameras. The dataset creation workflow encompasses raw data processing, data visualization methods, annotated data visualization methods, and other related steps. This dataset is primarily applied in the autonomous driving field to address challenges related to object recognition and environmental perception.