Novel Deep Learning Models Designed for Detecting Car Window Openings

As the level of automotive intelligence continues to advance, the window control system has become an essential component of automotive intelligence. This study aims to enhance the accuracy of window aperture detection, ensure user safety, and promote the development of automotive intelligence. We collected a dataset of window aperture under various scenarios and proposed a novel window aperture detection model named WiONet. The WiONet model integrates partial convolution and weighted feature fusion mechanisms to improve processing speed and feature fusion effectiveness. Through ablation experiments and comparisons with mainstream detection models, we verified the superiority of the WiONet model in terms of precision and F1-Score. Compared to the baseline model, the WiONet model achieved a 0.19 increase in precision and a 12% increase in F1-Score relative to other mainstream models. These results indicate that the WiONet model possesses higher accuracy in window aperture detection, providing users with a safer driving experience and offering new technical support for the digitalization and intelligence of the automotive industry.