The improper wearing or absence of helmets represents a significant contributing factor to fatal accidents in motorcycle driving. This dataset serves the purpose of detecting whether individuals have correctly or incorrectly worn helmets through camera-based analysis. The Helmet dataset has been curated, comprising a total of 28,736 images featuring various helmet types, including Full-Face, Half-Face, Modular, and Off-Road Helmets, in both correct and incorrect configurations. Captured using an iPhone 13 and Mi10T mobile phones, the images exhibit diverse climatic conditions, ranging from daytime to night-time scenarios. Subsequent to image acquisition, a pre-processing phase was undertaken to standardize the dataset. This involved renaming the images and adjusting their dimensions to a uniform 768 × 576 resolution, after which they were organized into respective folders. The uniqueness of this dataset lies in its incorporation of diverse environmental conditions, comprehensive helmet types, variability in helmet orientations, and its status as a large and balanced dataset, thereby presenting a realistic representation of real-world scenarios. The dataset's utility extends to various machine learning tasks, including image classification, object detection, and pose estimation specifically geared towards helmet recognition. Its scientific value lies in its potential to advance research and development in the realm of safety measures associated with motorcycle helmet usage.
Keywords: Bike; Helmet recognition; Image classification; Machine learning; Real-world scenarios; Safe driving.
© 2024 The Author(s).