IOD-Video
收藏资源简介:
IOD-Video数据集由南京大学创建,包含600个视频,总计141,017帧,涵盖多种距离、尺寸、可见度和场景,使用不同光谱范围捕捉。数据集主要用于非实体物体检测(IOD),这些物体具有无定形形状、模糊边界和缺乏颜色信息。数据集的创建过程涉及从数千个视频中筛选和标注,确保每帧都包含非实体物体。IOD-Video数据集的应用领域包括实时智能监控和早期预警,特别是在化学气体泄漏、火灾事故和有毒气体泄漏等场景中,旨在提高对这些难以识别物体的检测能力。
The IOD-Video dataset was developed by Nanjing University. It comprises 600 videos totaling 141,017 frames, covering diverse distances, object sizes, visibility conditions and scenarios, and was captured across different spectral ranges. Primarily designed for Intangible Object Detection (IOD), the dataset targets objects characterized by amorphous shapes, indistinct boundaries and a lack of color information. The dataset development process involved screening and annotating from thousands of source videos, ensuring that every frame contains intangible objects. The IOD-Video dataset is applied in real-time intelligent surveillance and early warning systems, particularly for scenarios such as chemical gas leaks, fire accidents and toxic gas leakage, with the goal of enhancing the detection performance for these hard-to-identify objects.




