RSOD
收藏资源简介:
RSOD数据集是由南京航空航天大学创建的高质量真实雪景对象检测数据集,包含2100张真实世界雪景图像,标注格式兼容MSCOCO和YOLO。该数据集专注于提升真实世界雪景场景下的对象检测精度,通过引入雪覆盖率(SCR)指标和独特的激活函数(Peak Act),将图像分为四个难度级别,以更好地理解雪对对象检测性能的影响。RSOD数据集适用于研究雪景下的对象检测问题,特别是在自动驾驶和监控等户外视觉系统中。
The RSOD dataset is a high-quality real-world snow scene object detection dataset created by Nanjing University of Aeronautics and Astronautics. It contains 2100 real-world snow scene images, with annotation formats compatible with MSCOCO and YOLO. This dataset focuses on improving the accuracy of object detection in real-world snow scenarios. By introducing the Snow Coverage Ratio (SCR) metric and a unique activation function (Peak Act), it divides images into four difficulty levels to better understand the impact of snow on object detection performance. The RSOD dataset is suitable for researching object detection in snow environments, especially for outdoor vision systems such as autonomous driving and surveillance.

- 1CF-YOLO: Cross Fusion YOLO for Object Detection in Adverse Weather with a High-quality Real Snow Dataset南京航空航天大学计算机科学与技术学院 · 2022年



