Rail-DB
收藏资源简介:
Rail-DB是由深圳技术大学创建的铁路检测数据集,包含7432对图像及其标注,涵盖多种光照、道路结构和视角条件。数据集中的轨道通过多边形进行标注,并根据背景被分为九种场景。Rail-DB旨在推动铁路检测算法的进步和比较,通过提供多样化的真实世界铁路图像,增强算法的鲁棒性。此外,数据集的创建过程包括从真实世界火车视频中获取图像,通过粗略和精细两个阶段进行标注,确保标注的准确性和完整性。Rail-DB的应用领域主要集中在铁路异常检测,特别是铁路区域的识别,以提高铁路安全和维护效率。
Rail-DB is a railway detection dataset developed by Shenzhen Technology University. It contains 7432 image pairs and their corresponding annotations, covering diverse lighting conditions, road structures, and viewing angles. The tracks in the dataset are annotated with polygons, and the images are categorized into nine scenarios based on their background features. Rail-DB aims to advance the development and comparative benchmarking of railway detection algorithms, enhancing the robustness of these algorithms by providing diverse real-world railway images. Furthermore, the dataset creation process involves extracting images from real-world train videos, with annotations completed in two stages: coarse and fine, to ensure the accuracy and completeness of the annotations. Its primary application domains focus on railway anomaly detection, particularly railway region recognition, to improve railway safety and maintenance efficiency.




