Learning to tell brake lights with convolutional features
Guangyu Zhong, Yi–Hsuan Tsai, Yi-Ting Chen, Xue Mei, Danil V. Prokhorov, Michael James, Ming–Hsuan Yang · 2016
In this paper, we present a learning-based brake light classification algorithm for intelligent driver-assistance systems. State-of-the-art approaches apply different image processing techniques with hand-crafted features to determine whether brake lights are on or off. In contrast, we learn a brake light classifier based on discriminative color descriptors and convolutional features fine-tuned for traffic scenes. We show how brake light regions can be segmented and classified in one framework. Numerous experimental results show that the proposed algorithm performs well against state-of-the-art alternatives in real-world scenes.