Rail-5k: a Real-World Dataset for Rail Surface Defects Detection

Zihao Zhang, Shaozuo Yu, Siwei Yang, Bingchen Zhao, Yu Zhou

Keywords: Real-world, Object detection, Semantic segmentation, Rail surface defects

TL;DR: We present a dataset for railway surface defects detection.

Abstract:

This paper presents the Rail-5k dataset for benchmarking the performance of visual algorithms in a real-world application scenario, namely the rail surface defect detection task.
We collected over 5k high-quality images from railways across China and annotated 1100 images with the help of railway experts to identify the most common 13 types of railway 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 make 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 tackled by previous semi-supervised learning methods.
We believe our dataset could be a valuable benchmark for evaluating the robustness and reliability of visual algorithms.

URL: https://www.dropbox.com/sh/yzq1g3asjz9a1kt/AAC3yNBE4W11lSEgjw2vqfpta?dl=0 , https://drive.google.com/drive/folders/1iJmWtjx0i2l_iwX48C29e6-_0lnnbUUs?usp=sharing

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2106.14366/code)