道路信息感知仿真图片

数据集的内容：共635张图片数据，共计144MB。包含：训练图片集500张，测试图片集100张，仿真图片集16张，特殊工况图片集19张。其中涵盖了车辆、行人、标志牌的标注数据，以及相关的测试和仿真数据。
采集方案：采用安装在车辆前挡风玻璃上的前视摄像头，分发给全国各地的采集志愿者，采用众包的方式采集原始视频，从原始视频中，筛选用于训练、测试和验证的数据。
采集地点：全国各地
设备情况：设备采用安霸A12平台安装前视1MP摄像头，图像分辨率1280*720，焦距5.47mm，HFOV52°，VFOV38.4°，Sensor采用AR0132AT, HDR>115dB

Dataset Content: A total of 635 image files with a total size of 144 MB, including 500 training images, 100 test images, 16 simulated images and 19 special working condition images. The dataset covers annotated data of vehicles, pedestrians and traffic signs, as well as corresponding test and simulated data.
Collection Scheme: A front-facing camera mounted on the front windshield of a vehicle was used. Raw videos were collected via crowdsourcing with volunteers across the country participating, and data for training, testing and validation were screened from these raw videos.
Collection Locations: Across the entire country.
Equipment Specifications: The system adopts the Ambarella A12 platform with a front-facing 1MP camera installed. The image resolution is 1280×720, focal length is 5.47 mm, HFOV is 52°, VFOV is 38.4°, the sensor is AR0132AT, and HDR performance exceeds 115 dB.

搜集汇总

背景与挑战

背景概述

该数据集包含635张道路信息感知仿真图片（144MB），分为训练、测试、仿真和特殊工况四个子集，涵盖车辆、行人、标志牌的标注数据。数据通过众包方式采集，使用前视摄像头在全国各地拍摄，图像分辨率为1280*720，具有较高的动态范围（HDR>115dB）。