Rail-5k: a Real-World Dataset for Rail Surface Defects Detection
This provides a new dataset for researchers and practitioners in railway maintenance to evaluate visual algorithms, but it is incremental as it focuses on data creation rather than novel methods.
The paper introduces the Rail-5k dataset, a real-world collection of over 5,000 images for rail surface defect detection, with 1,100 labeled images covering 13 defect types, to benchmark visual algorithms in fully-supervised and semi-supervised settings.
This paper presents the Rail-5k dataset for benchmarking the performance of visual algorithms in a real-world application scenario, namely the rail surface defects detection task. We collected over 5k high-quality images from railways across China, and annotated 1100 images with the help from railway experts to identify the most common 13 types of rail defects. The dataset can be used for two settings both with unique challenges, the first is the fully-supervised setting using the 1k+ labeled images for training, fine-grained nature and long-tailed distribution of defect classes makes it hard for visual algorithms to tackle. The second is the semi-supervised learning setting facilitated by the 4k unlabeled images, these 4k images are uncurated containing possible image corruptions and domain shift with the labeled images, which can not be easily tackle by previous semi-supervised learning methods. We believe our dataset could be a valuable benchmark for evaluating robustness and reliability of visual algorithms.