Paper Title
Motorcycle Helmet Dataset: An Image Based Dataset for Helmet Detection of Bike Riders

Abstract
Existing object detection algorithms achieve high image recognition performance with good lighting and proximity to the imaging device. However, most of the existing algorithms fail to perform equally well in viewing situations, where images are captured across varying resolutions and spectra. In surveillance settings, cameras are often placed far away from the subjects, thus leading to a change in orientation, illumination, occlusion, and resolution. Current datasets used for helmet detection are captured in constrained environments, and as a result, fail to mimic real-world conditions. In this paper, we present the MotorcycleHelmetDataset that contains 4 different subjects in a total of 1460 images. The proposed database contains more than 1460 images containing Bike Rider, WithHelmet and WithoutHelmet, which are distributed across images captured in visible light. Each image is captured when the bike rider is nearly 30ft from the imaging device, offering a host of common challenges in viewing settings. Keywords - Image recognition, LabelImg, Object detection.