Real–Time Vehicle Detection Based on Wavelet Decomposition and CNN

This paper presents a new approach for real-time vehicle detection and classification using a fully convolutional neural network (FCNN) with wavelet transform as a preprocessing technique. Real-time prediction is crucial for efficient traffic control, as it enables timely decision-making and response. The proposed approach addresses this issue by using stacked images obtained from wavelet transform as input to the FCNN, resulting in faster inference time and lower number of FLOPs compared to traditional CNN-based methods. Additionally, the proposed method shows promising results in terms of accuracy, especially with some variations of the base model. Overall, this approach represents a significant step forward in the development of real-time object detection systems.