LLVIP
收藏资源简介:
LLVIP是由北京邮电大学创建的一个可见光-红外配对数据集,专为低光视觉任务设计。该数据集包含30976张图像,总计15488对,主要在极暗场景下拍摄,所有图像在时间和空间上严格对齐。数据集中的行人已被标注,适用于图像融合、行人检测和图像到图像的转换等任务。创建过程中,使用双目相机收集图像,并通过注册确保可见光和红外图像的对齐。LLVIP数据集的应用领域包括提升低光条件下的图像融合、行人检测和图像转换算法性能,以及支持多模态图像注册和领域适应研究。
LLVIP is a visible-infrared paired dataset developed by Beijing University of Posts and Telecommunications, tailored specifically for low-light vision tasks. It contains 30,976 images, amounting to 15,488 image pairs, and was primarily captured in extremely dark scenes. All images are strictly aligned both temporally and spatially. Pedestrians in the dataset have been annotated, making it suitable for tasks such as image fusion, pedestrian detection, and image-to-image translation. During data collection, a binocular camera was used, and the alignment between visible and infrared images was ensured through registration. The application scenarios of the LLVIP dataset include improving the performance of image fusion, pedestrian detection, and image translation algorithms under low-light conditions, as well as supporting research on multimodal image registration and domain adaptation.




