What question did this study set out to answer?

The research aims to enhance detection, tracking, and visualization of transport entities in urban environments using a hybrid deep learning framework.

June 10, 2026Open Access

A Deep Learning Approach for AI-Based Detection and Visualization in Smart Transportation Systems Using CNN and SNN models

YMYasothapriya MAvila University BRBharathidasan RAvila University

Key Points

The research aims to enhance detection, tracking, and visualization of transport entities in urban environments using a hybrid deep learning framework.
Proposed a hybrid model integrating CNNs and SNNs for image analysis and classification
Evaluated on benchmark traffic datasets and an Indian smart city dataset
Implemented an interactive AI-based visualization layer for real-time traffic flow modeling
Achieved Mean Average Precision (mAP) of 96.8% and accuracy of 98.4%
Reduced energy consumption by 43% compared to traditional CNN architectures
Demonstrated effectiveness in real-time detection and visualization for Intelligent Transportation Systems

Abstract

The rapid urbanization and exponential growth of vehicular traffic have necessitated the evolution of traditional traffic management systems into Artificial Intelligence (AI)-driven Smart Transportation Systems (STS). In real-world urban environments, achieving real-time detection, tracking, and visualization of dynamic transport entities poses a monumental challenge due to varying lighting conditions, severe occlusions, sensor noise, and computational bottlenecks at edge devices. This article proposes a novel hybrid deep learning framework that synergistically integrates Convolutional Neural Networks (CNNs) and Spiking Neural Networks (SNNs) for highly accurate and energy-efficient AI-based detection and real-time visualization in STS. The proposed hybrid CNN-SNN architecture leverages the profound spatial feature extraction capabilities of CNNs to analyze complex image data acquired from smart city cameras. Simultaneously, it maps these deep spatial hierarchies into the temporal and energy-efficient event-driven processing paradigm of SNNs. The SNN layer serves as an asynchronous spiking classifier that drastically reduces inference latency and power consumption, rendering the model suitable for resource-constrained edge computing environments. Furthermore, a highly interactive AI-based visualization layer is introduced to translate multidimensional detection outputs into real-time traffic flow models and anomaly spatial maps. Extensive experiments were conducted evaluating the framework on leading benchmark traffic datasets as well as a locally acquired Indian smart city dataset. The proposed framework demonstrates a marked improvement over state-of-the-art models, achieving a Mean Average Precision (mAP) of 96.8% and accuracy of 98.4%, while reducing energy consumption by 43% compared to conventional deeply layered CNN architectures. The findings underscore the efficacy of hybrid CNN-SNN models as a robust, scalable, and sustainable solution tailored for the next generation of Intelligent Transportation Systems

Ask AI

Helpful

Bookmark

View Full Paper