Developing an AI System for Smart Traffic: From Vehicle Detection to Data Analytics Dashboard
Project trailer: transcript (3 min 23 s)
I-Wei Huang and Yu-Hsi Li, advised by Prof. Gary Shueng-Han Chan
Department of Computer Science and Engineering, The Hong Kong University of Science and Technology

[0:00] Title card: "Developing an AI System for Smart Traffic: From Vehicle Detection to Data Analytics Dashboard", I-Wei Huang and Yu-Hsi Li, advised by Prof. Gary Shueng-Han Chan.

[0:07] Night-time traffic in Hong Kong, with on-screen text: "Hong Kong is one of the most densely populated cities in the world. Studies show that traffic congestion increases commute time by over 70% during peak hours."

[0:18] Narrator: That's not just inconvenient. It's costly in terms of time, efficiency, and experience.

[0:25] K11 MUSEA, "a premier shopping and cultural complex located in the heart of Victoria Dockside, Tsim Sha Tsui".

[0:30] Narrator: On weekends and public holidays, one of its key surrounding roads, Salisbury Road, turns into a bottleneck. You'll see buses lined up at the stop, taxis crowding the drop-off zones, and private cars queuing to enter the car park. While this influx of visitors creates a vibrant atmosphere, it also brings frustration for drivers and operational headaches for the K11 authority.

[0:52] Narrator: Given this recurring congestion, K11 has recognized the urgent need for actionable insights to direct traffic flow. They need to understand what's happening outside their doors in real time, so they can respond quickly, improve traffic, and elevate visitor experience.

[1:09] Narrator: In response to these challenges, the goal of our project, developed in collaboration with K11, aims to deliver a real-time smart traffic data analytics system.

[1:19] Narrator: Our system leverages computer vision technology to deliver three layers of actionable insights. Firstly, we accurately extract live traffic metrics like vehicle count and speed to help K11 manage traffic around entrances and redirect flow when needed. Secondly, we capture snapshots of illegally parked vehicles and track vehicle type composition, enabling fast, targeted intervention. Lastly, we provide historical insights for trend analysis to proactively deploy traffic guides or adjust flow strategies.
On screen: "Dynamic Entrance Guidance", "Proactive Staff Deployment", "Data-Driven Traffic Planning".

[1:50] Narrator: Our system includes a camera with an edge traffic analyzer and a dashboard visualization. The edge traffic analyzer transforms the raw video footage into critical traffic insight and sends it to the dashboard server.
On screen: a diagram of Camera, Edge Traffic Analyzer, and Dashboard.

[2:03] Narrator: This is how the actual system looks.

[2:08] Screen recording, "Zone Setup": by clicking, select the zone area to be analyzed. Two zones are drawn over the camera view of Salisbury Road.

[2:34] Screen recording, "Program Start" and "Dashboard": the analyzer starts and sends data to the dashboard in real time ("Real-time data transfer"). The dashboard shows the location map, congestion level, flow rate, vehicle count over time, average speed, vehicle types, and the latest traffic snapshot.

[3:09] On screen: "Real-Time", "Actionable Insights", "Scalable".
Narrator: Our system delivers real-time updates, offers actionable insights, and is scalable for diverse locations. We empower decision-makers with smart, efficient, and proactive traffic management tools.

Note: this transcript follows the subtitles shown in the video. Licence plates in the stock footage between 0:06 and 0:53 are blurred in the version published on this page.