AbstractVisual inspection is one of the main approaches for annual bridge inspection. Light detection and ranging (LiDAR) scanning is a new technology, which is beneficial because it collects the point clouds and the third dimension of the scanned objects. Deep learning (DL)-based methods have attracted researchers’ attention for concrete surface defect detection. However, no point cloud–based DL method currently is available for semantic segmentation of bridge surface defects without converting the data set into other representations, which results in increasing the size of the data set. Moreover, most of the current point cloud–based concrete surface defect detection methods focus on only one type of defect. On the other hand, a data set plays a key role in DL. Therefore, the lack of publicly available point cloud data sets for bridge surface defects is one of the reasons for the lack of studies in this area. To address these issues, this paper created a publicly available point cloud data set for concrete bridge surface defect detection, and developed a point cloud–based semantic segmentation DL method to detect different types of concrete surface defects. Surface Normal Enhanced PointNet++ (SNEPointNet++) was developed for semantic segmentation of concrete bridge surface defects (i.e., cracks and spalls). SNEPointNet++ focuses on two main characteristics related to surface defects (i.e., normal vector and depth) and considers the issues related to the data set (i.e., imbalanced data set). The data set, which was collected from four concrete bridges and classified into three classes (cracks, spalls, and no defect), is made available for other researchers. The model was trained and evaluated using 60% and 20% of the data set, respectively. Testing on the remaining part of the data set resulted in 93% and 92% recall for cracks and spalls, respectively. Spalls of the segments deeper than 7 cm (severe spalls) can be detected with 99% recall.