Summon Auntie

Problem

Current enforcement efforts rely heavily on in-person patrols (e.g. routine checks, responding to complaints, or targeted operations at potential hotspots). This means that violations could go undetected or face delayed detection.

While there are some permanent and temporary camera deployments, these systems are typically single-purpose – designed to detect only one type of violation.

🚘 How might we detect vehicle-related offences more rapidly and comprehensively, to create safer and smoother roads for all?

Product Goals

We want to create an enforcement system that is:

• Accurate: Meets or exceeds the accuracy of current manual enforcement efforts and camera deployments.

• Cost-efficient: Results in manpower and cost savings for agencies.

• Scaleable: Able to detect different types of offences and be deployed to different geographical areas.

Our Approach

1. Understand: the landscape of traffic and vehicle-related offences in Singapore, how enforcement is currently done, what is important to the relevant agencies, and what our existing ANPR cameras can/cannot do.

2. Scope down: to what offences we think are most impactful to focus on and what we can do in 1 month.

Key Features of our Prototype:

Real-time processing of camera data

Our ANPR cameras are cloud-connected so vehicle detections are streamed directly to secure servers for immediate analysis.

Location of cameras we used for our prototype

Detection of off-peak vehicles during restricted hours and lorries with inadequate safety measures

Each offence has a custom detection logic. For more complex violations, we use AI computer vision.

Admin portal to verify and export the details of verified violations

Our assumption is that any follow-up actions are done in a separate system.

What's Next?

🚍 Expand our prototype to cover other offences (what offences do you think we should cover next? )

💬 Get feedback from relevant agencies.