Blind spot detection is an important feature of Advanced Driver Assistance Systems (ADAS).\nIn this paper, we provide a camera-based deep learning method that accurately detects other vehicles\nin the blind spot, replacing the traditional higher cost solution using radars. The recent breakthrough\nof deep learning algorithms shows extraordinary performance when applied to many computer\nvision tasks. Many new convolutional neural network (CNN) structures have been proposed and\nmost of the networks are very deep in order to achieve the state-of-art performance when evaluated\nwith benchmarks. However, blind spot detection, as a real-time embedded system application,\nrequires high speed processing and low computational complexity. Hereby, we propose a novel\nmethod that transfers blind spot detection to an image classification task. Subsequently, a series of\nexperiments are conducted to design an efficient neural network by comparing some of the latest\ndeep learning models. Furthermore, we create a dataset with more than 10,000 labeled images using\nthe blind spot view camera mounted on a test vehicle. Finally, we train the proposed deep learning\nmodel and evaluate its performance on the dataset.
Loading....