PANDA
收藏资源简介:
PANDA是一个具有高分辨率的人类中心视频数据集,由清华大学创建,旨在支持大规模、长期和多对象的视觉分析。该数据集通过高分辨率相机捕捉真实世界场景,覆盖广阔的视野和高空间分辨率。PANDA包含丰富的层次化标注,包括15,974.6k边界框、111.8k细粒度属性标签、12.7k轨迹、2.2k群体和2.9k交互。数据集的应用领域包括人群行为分析和交互检测,旨在解决复杂人群行为和交互的识别问题。
PANDA is a high-resolution human-centric video dataset developed by Tsinghua University, which aims to support large-scale, long-term and multi-object visual analysis. This dataset captures real-world scenes via high-resolution cameras, boasting wide field of view and high spatial resolution. PANDA includes rich hierarchical annotations, namely 15,974.6k bounding boxes, 111.8k fine-grained attribute labels, 12.7k trajectories, 2.2k groups and 2.9k interactions. Its application domains cover crowd behavior analysis and interaction detection, targeting the recognition of complex crowd behaviors and interactions.
- 1PANDA: A Gigapixel-level Human-centric Video Dataset清华大学 · 2020年



