vapaau/LTDv2

LTDv2数据集是一个大规模长期热漂移监控多目标检测数据集，包含超过100万帧图像，时间跨度为8个月。数据集中有超过680万个标注边界框，涵盖4个对象类别（人、自行车、摩托车、车辆）。该数据集跨越多个季节、天气条件和昼夜周期，目标和背景随时间展示出显著的外观变化。每张图像还配有关于该地区天气相关环境条件的元数据，为视觉域提供辅助信息。

The LTDv2 dataset is a large-scale long-term thermal drift surveillance multi-object detection dataset consisting of over 1 million frames extracted from video clips spanning a period of 8 months. The dataset contains more than 6.8 million annotated bounding boxes across 4 object classes (Person, Bicycle, Motorcycle, Vehicle). Encompassing several seasons, weather conditions, and day-night cycles, the objects and background exhibit drastic variations in appearance over time. Each image is also accompanied by metadata detailing the weather-related environmental conditions in the area, providing supplementary information to aid the visual domain.