ParkingSticker

ParkingSticker是一个模拟工业问题的真实世界物体检测数据集，由杜克大学创建。该数据集包含1871张来自安全摄像头视频片段的图像，旨在识别接近安全摄像头门禁的汽车上的停车贴纸。停车贴纸在图像中用边界框标记，平均尺寸远小于其他流行物体检测数据集中的物体，增加了检测难度。数据集真实反映了工业问题中客户提供的几帧视频并要求解决非常困难问题的场景。该数据集可用于评估不同物体检测方法的性能，并帮助研究人员解决具有真实世界约束的实际问题，如非理想摄像头定位和小物体-图像尺寸比。ParkingSticker is a real-world object detection dataset simulating industrial scenarios, developed by Duke University. This dataset contains 1871 images extracted from security camera footage, aiming to identify parking stickers on vehicles approaching security camera access control points. Parking stickers in these images are annotated with bounding boxes. Their average size is considerably smaller than objects in other mainstream object detection datasets, which elevates the difficulty of detection. This dataset realistically reflects the scenario where clients provide several video frames from industrial settings and require solutions to highly challenging tasks. It can be used to evaluate the performance of various object detection methods, and assist researchers in addressing real-world practical problems with constraints such as non-ideal camera placement and small object-to-image size ratios.