DistilBrakeLightNet: Knowledge Distilled Approach for Classification of Car Brake Light

The ability to accurately detect if brake lights are ON or OFF in vehicles is a crucial component in advanced driver assistance systems (ADAS) and autonomous driving technologies, as it allows for improved situational awareness and enhanced safety features. This paper presents a novel approach to the brake light detection using a knowledge distilled network that’s derived from the well-known VGG16 architecture. The proposed method leverages the powerful ‘feature extraction capabilities’ of the VGG16 model, while significantly reducing the computational complexity through network distillation. This paper aims to implement the distilled network, which is shallow as compared to standard network VGG16, while delivering the accuracy almost equal to VGG16. The paper also aims to run the networks such as MobileNet and ResNet50 to compare their metrics with distilled network.