Back to Publications

Author(s) Lin, H., Chen, G, Siebert, F. W.
Title Positional Encoding: Improving class-imbalanced motorcycle helmet use classification
Abstract Recent advances in the automated detection of motorcycle riders’ helmet use have enabled road safety actors to process large scale video data efficiently and with high accuracy. To distinguish drivers from passengers in helmet use, the most straightforward way is to train a multi-class classifier, where each class corresponds to a specific combination of rider po- sition and individual riders’ helmet use. However, such strat- egy results in long-tailed data distribution, with critically low class samples for a number of uncommon classes. In this pa- per, we propose a novel approach to address this limitation. Let n be the maximum number of riders a motorcycle can hold, we encode the helmet use on a motorcycle as a vector with 2n bits, where the first n bits denote if the encoded posi- tions have riders, and the latter n bits denote if the rider in the corresponding position wears a helmet. With the novel helmet use positional encoding, we propose a deep learning model that stands on existing image classification architecture. The model simultaneously trains 2n binary classifiers, which al- lows more balanced samples for training. This method is sim- ple to implement and requires no hyperparameter tuning. Ex- perimental results demonstrate our approach outperforms the state-of-the-art approaches by 1.9% accuracy.