Parking Pedestrian Dataset (PPD)
收藏资源简介:
停车行人数据集(PPD)是一个大规模的鱼眼数据集,专注于真实世界中行人的检测,特别是在遮挡和多样姿态的情况下。PPD包含多种类型的行人,通过鱼眼摄像头捕捉,旨在解决停车场场景中行人检测的挑战。数据集由复旦大学和纵目科技创建,包含超过33万张鱼眼图像,分为遮挡行人、姿态行人和躺下行人三个子集。PPD通过引入两种数据增强技术(ODA和PDA)来提高行人检测的性能,适用于自动驾驶中的高级驾驶辅助系统(ADAS)研究。
The Parking Pedestrian Dataset (PPD) is a large-scale fisheye dataset dedicated to real-world pedestrian detection, especially in scenarios with occlusions and diverse poses. The PPD captures various types of pedestrians via fisheye cameras, aiming to address the challenges of pedestrian detection in parking lot scenarios. Developed by Fudan University and Zongmu Technology, the dataset contains over 330,000 fisheye images, which are divided into three subsets: occluded pedestrians, pedestrians with diverse poses, and lying pedestrians. Moreover, PPD introduces two data augmentation techniques, ODA and PDA, to improve pedestrian detection performance, making it suitable for research on Advanced Driver Assistance Systems (ADAS) in autonomous driving.

- 1PPD: A New Valet Parking Pedestrian Fisheye Dataset for Autonomous Driving复旦大学和纵目科技 · 2023年



