evCIVIL : Event-Based Dataset for visual Detection of CIVIL infrastructure Defects

evCIVIL dataset is dataset which is designed for the purpose of civil infrastructure defect detection using event-based cameras(DVS cameras). evCIVIL contains both asynchronous event streams and intensity gray scale images captured simultaneously with DAVIS346 camera. The dataset concentrates majorly on two types of civil infrastructure defects : cracks and spalling. These defects were captured from various types of civil structures such as roads, buildings, pavements, tunnels and so on under different lighting conditions and weather conditions.The field dataset comprises 318 recording sequences which covers about 80 minutes, documenting 458 distinct cracksand 121 distinct spalling instances. The laboratory dataset includes 362 recording sequences, covering 220 distinctcracks and 308 spalling instances.For a given sequence the captured events and gray scales images were available in seperate .h5 files. The defects are labelled as bounding boxes together with their defect categories (whether crack or spalling defect) and those information is provided in ".npy" file.For a comprehensive and in detailed description of the data collection process, dataset and dataset benchmarking please refer to our paper "Event-based Civil Infrastructure Visual Defect Detection: ev-CIVIL Dataset and Benchmark".For all the code that used to preprocess and deploy dataset , dataset usage and benchmarking please refer to our github repository (https://github.com/gwgknudayanga/evCIVIL).