JHU-CROWD
收藏资源简介:
JHU-CROWD是由约翰霍普金斯大学电气与计算机工程系创建的大规模无约束人群计数数据集,包含4,250张图像和111万个标注。该数据集涵盖了多种场景和环境条件,特别包括了受天气影响的退化图像和无人群的干扰图像,使其成为一个极具挑战性的数据集。数据集不仅提供了图像级别的丰富标注,还包括头部级别的详细信息,如遮挡、模糊和大小等。JHU-CROWD旨在解决现有数据集在多样性和标注丰富性方面的不足,为人群计数技术的发展提供更全面的基准。
JHU-CROWD is a large-scale unconstrained crowd counting dataset developed by the Department of Electrical and Computer Engineering at Johns Hopkins University, which consists of 4,250 images and over 1.11 million annotations. This dataset covers diverse scenarios and environmental conditions, specifically including weather-degraded images and crowd-free distractor images, making it an extremely challenging benchmark. It not only provides rich image-level annotations, but also includes detailed head-level information such as occlusion, blurriness, and head size. JHU-CROWD aims to address the limitations of existing datasets in terms of diversity and annotation richness, providing a more comprehensive benchmark for the advancement of crowd counting techniques.
- 1Pushing the Frontiers of Unconstrained Crowd Counting: New Dataset and Benchmark Method约翰霍普金斯大学电气与计算机工程系 · 2019年



