EgoSDQES
收藏资源简介:
EgoSDQES数据集是由斯坦福大学创建的一个新基准,基于Ego4D数据集,旨在支持流式检测查询事件开始(SDQES)任务。该数据集包含大量从第一人称视角拍摄的长视频,涵盖多样化的活动、视角和相机运动,适用于评估模型在复杂现实场景中的鲁棒性。数据集的创建过程涉及从Ego4D数据集中提取并注释自然语言查询,以捕捉复杂事件的开始。EgoSDQES数据集主要应用于机器人、自动驾驶和增强现实等需要实时反应的领域,旨在解决复杂事件的低延迟检测问题。
The EgoSDQES dataset is a novel benchmark developed by Stanford University based on the Ego4D dataset, aiming to support the Streaming Detection of Query Event Start (SDQES) task. This dataset includes a large number of long videos captured from first-person perspectives, covering diverse activities, viewpoints and camera movements, and is applicable to evaluating the robustness of models in complex real-world scenarios. The creation of the EgoSDQES dataset involves extracting and annotating natural language queries from the Ego4D dataset to capture the start of complex events. The EgoSDQES dataset is mainly used in fields requiring real-time response such as robotics, autonomous driving and augmented reality, with the goal of solving the low-latency detection problem of complex events.

- 1Streaming Detection of Queried Event Start斯坦福大学 · 2024年



