The git provides a dataset for a safety helmet classifier, which is based on Datafountain’s dataset. The DataFountain’s dataset can be downloaded from https://www.datafountain.cn/competitions/304/datasets after registering. It is a real dataset for safety helmet detection since it has the ground-truth labels for all safety helmets in the image set. The Train_dataset includes all positive and negative samples for training a safety helmet detector, and the Test_dataset includes all positive and negative samples for testing the safety helmet detector. Moreover, we constructed a face-detection-based safety helmet detector. The corresponding paper has been submitted to Computer-Aided Civil and Infrastructure Engineering. The idea is in the following figure. The raw_images provides some images from various construction sites to verify a safety helmet detector. The directory raw_images only stores parts of images, and you can download all images from https://drive.google.com/open?id=1Jn5UcaQ0fB9WDow5oiBUXiUhkSQf0YSo.
