WiderPerson

WiderPerson总共包含13,382个带有399,786注释的图像，即每个图像29.87个注释，这意味着该数据集包含具有各种遮挡的密集行人。因此，由于场景和遮挡的差异很大，因此建议的数据集中的行人极具挑战性，这适合于评估野外的行人检测器。WiderPerson comprises a total of 13,382 images with 399,786 annotations, amounting to 29.87 annotations per image. This indicates that the dataset contains dense pedestrians with various occlusions. Given the substantial variations in both scene contexts and occlusion scenarios, pedestrians in this dataset are highly challenging, making it suitable for evaluating pedestrian detectors in the wild.