复杂交通道路的自动驾驶数据

采集杭州城市或乡村道路相机图像和点云数据，对多个场景（白天、阴天、雨天、夜晚、拥堵等）的道路数据进行采集。通过一定的算法规则对数据进行处理后，用数据训练道路拥堵算法模型，反映道路拥堵情况，从而指导采取措施缓解道路拥堵情况。1.首先对收集的相机图像和点云数据进行了等间隔采样，其次，针对图像数据，使用预训练的2D/3D检测模型进行预检测实验，对目标数极少及运动模糊的数据进行滤除。2.最终得到：场景丰富的交通图像与点云数据。3.解决问题：采样处理避免由于时间戳密度大而降低数据的多样性，预训练模型剔除了大量冗余和低质数据。

This dataset collects camera images and point cloud data from urban and rural roads in Hangzhou, covering diverse scenarios including daytime, overcast, rainy, nighttime, and traffic congestion. After processing the acquired data with specified algorithmic rules, the processed data is used to train road congestion detection models to reflect road congestion conditions, thereby guiding the implementation of measures to alleviate road congestion.
1. First, equally-spaced sampling is performed on the collected camera images and point cloud data. Subsequently, pre-detection experiments are conducted on the image data using pre-trained 2D/3D detection models, and data with extremely few target objects and motion blur is filtered out.
2. Ultimately, a dataset of traffic images and point cloud data with rich and diverse scenarios is obtained.
3. The processing addresses the following issues: the equally-spaced sampling avoids reducing data diversity caused by overly dense timestamps, while the pre-trained models eliminate a large amount of redundant and low-quality data.

搜集汇总

数据集介绍

特点

该数据集是杭州云象网络技术有限公司提供的复杂交通道路的自动驾驶数据，包含1360条每日更新的图像和点云记录，用于训练道路拥堵算法模型，覆盖多种天气和交通场景。