Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Real-Time Pedestrian Detection on IoT Edge Devices: A Lightweight Deep Learning Approach

Muhammad Dany Alfikri, Rafael Kaliski

TL;DR

The paper tackles real-time pedestrian detection on resource-constrained IoT edge devices by designing a lightweight, edge-optimized detector based on Tiny-YOLOv3 with a MobileNetV2 backbone. It deploys the model on an Nvidia Jetson Nano and uses MQTT for low-bandwidth, real-time event streaming to an edge server, achieving fast inference (~147 ms) and practical frame rates with around 78% accuracy. The study compares the proposed model against larger DL detectors, showing substantially lower memory and compute requirements while maintaining competitive performance, and demonstrates a full edge-to-edge workflow from camera to MQTT broker. The results suggest a viable, low-cost, real-time pedestrian detection pipeline for intelligent transportation systems, with future work focusing on cellular network integration and robustness enhancements.

Abstract

Artificial intelligence (AI) has become integral to our everyday lives. Computer vision has advanced to the point where it can play the safety critical role of detecting pedestrians at road intersections in intelligent transportation systems and alert vehicular traffic as to potential collisions. Centralized computing analyzes camera feeds and generates alerts for nearby vehicles. However, real-time applications face challenges such as latency, limited data transfer speeds, and the risk of life loss. Edge servers offer a potential solution for real-time applications, providing localized computing and storage resources and lower response times. Unfortunately, edge servers have limited processing power. Lightweight deep learning (DL) techniques enable edge servers to utilize compressed deep neural network (DNN) models. The research explores implementing a lightweight DL model on Artificial Intelligence of Things (AIoT) edge devices. An optimized You Only Look Once (YOLO) based DL model is deployed for real-time pedestrian detection, with detection events transmitted to the edge server using the Message Queuing Telemetry Transport (MQTT) protocol. The simulation results demonstrate that the optimized YOLO model can achieve real-time pedestrian detection, with a fast inference speed of 147 milliseconds, a frame rate of 2.3 frames per second, and an accuracy of 78%, representing significant improvements over baseline models.

Real-Time Pedestrian Detection on IoT Edge Devices: A Lightweight Deep Learning Approach

TL;DR

The paper tackles real-time pedestrian detection on resource-constrained IoT edge devices by designing a lightweight, edge-optimized detector based on Tiny-YOLOv3 with a MobileNetV2 backbone. It deploys the model on an Nvidia Jetson Nano and uses MQTT for low-bandwidth, real-time event streaming to an edge server, achieving fast inference (~147 ms) and practical frame rates with around 78% accuracy. The study compares the proposed model against larger DL detectors, showing substantially lower memory and compute requirements while maintaining competitive performance, and demonstrates a full edge-to-edge workflow from camera to MQTT broker. The results suggest a viable, low-cost, real-time pedestrian detection pipeline for intelligent transportation systems, with future work focusing on cellular network integration and robustness enhancements.

Abstract

Artificial intelligence (AI) has become integral to our everyday lives. Computer vision has advanced to the point where it can play the safety critical role of detecting pedestrians at road intersections in intelligent transportation systems and alert vehicular traffic as to potential collisions. Centralized computing analyzes camera feeds and generates alerts for nearby vehicles. However, real-time applications face challenges such as latency, limited data transfer speeds, and the risk of life loss. Edge servers offer a potential solution for real-time applications, providing localized computing and storage resources and lower response times. Unfortunately, edge servers have limited processing power. Lightweight deep learning (DL) techniques enable edge servers to utilize compressed deep neural network (DNN) models. The research explores implementing a lightweight DL model on Artificial Intelligence of Things (AIoT) edge devices. An optimized You Only Look Once (YOLO) based DL model is deployed for real-time pedestrian detection, with detection events transmitted to the edge server using the Message Queuing Telemetry Transport (MQTT) protocol. The simulation results demonstrate that the optimized YOLO model can achieve real-time pedestrian detection, with a fast inference speed of 147 milliseconds, a frame rate of 2.3 frames per second, and an accuracy of 78%, representing significant improvements over baseline models.
Paper Structure (20 sections, 10 figures, 3 tables)

This paper contains 20 sections, 10 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Research Contributions
  3. Research Organization
  4. Related Works
  5. Handcrafted Features based Pedestrian Detection
  6. Deep Learning based pedestrian detection
  7. System Model and Proposed Methods
  8. Proposed pedestrian Model for edge computing
  9. Deep Learning Model
  10. Model Backbone
  11. Model Neck and Head
  12. Dataset
  13. MQTT protocol for Internet of Things
  14. Simulation Setup
  15. Deep Learning Model
  16. ...and 5 more sections

Figures (10)

  • Figure 1: Traffic light and intersection system model
  • Figure 2: Geometric diagram of camera deployment
  • Figure 3: Camera placement diagram
  • Figure 4: YOLOv3-based deep learning model architecture
  • Figure 5: Our optimized Tiny YOLO V3 model architecture
  • ...and 5 more figures