minkyuchoi/Temporal-Logic-Video-Dataset

Temporal Logic Video (TLV)数据集旨在解决现有视频数据集在长时间跨度和时间扩展活动及物体检测方面的不足。该数据集由两部分组成：1. 合成数据集，通过拼接COCO和ImageNet的静态图像生成，引入各种时间逻辑规范；2. 真实世界数据集，基于开源自动驾驶数据集NuScenes和Waymo，提供真实世界的自动驾驶场景，并添加时间逻辑标注。数据集以序列化对象的形式提供，每个对象包含一组带有标注的帧。

The Temporal Logic Video (TLV) Dataset is a synthetic and real video dataset with temporal logic annotation. It addresses the scarcity of state-of-the-art video datasets for long-horizon, temporally extended activity and object detection. The dataset includes synthetic datasets (generated by stitching static images from COCO and ImageNet, introducing various temporal logic specifications) and real-world datasets (based on NuScenes and Waymo autonomous vehicle datasets, adding temporal logic specifications). The dataset structure includes serialized objects, each containing a set of annotated frames. Dataset statistics include the total number of frames and datasets.