Flowsense

Inspiration

A mobility decision can shape a city for years. Congestion consumes fuel, raises transport costs and reduces the capacity of businesses and public services. Governments must decide which junction needs attention, which corridor deserves investment, and how to preserve access to work, goods and essential services. Those choices compete for limited public budgets and cross departmental boundaries.

We built FlowSense around a practical question: how can a city turn a roadway observation into a reviewable decision, a responsible owner and a measurable outcome? Our starting point is a shared evidence workflow that helps mobility teams coordinate action and build a stronger basis for public investment.

The scale is tangible: the Los Angeles-Long Beach-Anaheim urban area alone carried an estimated $29.5 billion in annual congestion costs in 2024. That is one example of the stakes behind better mobility decisions. Texas A&M Transportation Institute, 2025 Urban Mobility Report.

What it does

FlowSense brings camera observations, road-network context and case review into one mobility workspace. Operators can inspect roadway evidence, review cases, compare route alternatives and retain the context behind a decision.

Its proposed government pilot connects mobility, infrastructure, transport and emergency operations around a shared corridor review. Consider a recurring freight bottleneck: mobility reviews the observations, transport examines the affected corridor, and infrastructure uses the evidence to scope an intervention. Emergency operations can also review road access when relevant. Authorized personnel retain operational decisions.

The wider ambition is to help governments identify recurring access problems, prioritize interventions and evaluate what changes afterward. Camera evidence contributes to that decision alongside engineering, demand and cost studies.

Our ambition aligns with SDG 9 through evidence for resilient infrastructure, SDG 11 through urban mobility and access, and SDG 13 through evaluation of interventions aimed at reducing avoidable fuel consumption and emissions.

How we built it

OpenCV 5 DNN runs a YOLOv8n model exported to ONNX to produce object detections and instantaneous class counts from camera images. FastAPI connects those observations to a React and TypeScript interface with Leaflet maps. OpenStreetMap, OSMnx and NetworkX provide the road-network foundation for comparing route alternatives.

We organized these components around the operator's decision: inspect the source, review the evidence, compare options and preserve the decision history. This connects computer vision to a concrete public-sector workflow.

Challenges

The central design challenge was making an observation useful to several public teams with different responsibilities. We chose a shared review workflow with source context, explicit ownership and a retained decision history.

We also prioritized visible source status and clear distinctions between observations and illustrative analytics. These decisions help operators judge the evidence before acting and give the next validation stage a practical focus: dependable inputs, connected workflows and reproducible deployment.

Accomplishments

Our October 15 verification recorded 91 passing software tests, successful frontend build and lint checks, and a public Caltrans frame processed through OpenCV 5.0.0 DNN. Interface checks confirmed Overview rendering and navigation to Risk Signals.

The repository brings case review, exports and graph-based route alternatives into one product. These are concrete building blocks for evaluating a government workflow from observation to reviewed action.

What we learned

An observation becomes useful when someone can explain the decision it supports. City teams need to know what was observed, who reviewed it, which action was chosen and how the result will be evaluated.

That lesson shaped FlowSense around evidence, ownership and follow-through. A clear operational record can serve today's review and inform tomorrow's investment discussion.

What's next

Take the evidence-to-decision workflow into a focused corridor pilot with a lead mobility authority and relevant partner teams. Establish baseline measures for verification time, case handling and road-access continuity. Evaluate a selected intervention with suitable before-and-after measurements, accounting for other changes.

The next stage is to establish which decisions FlowSense improves and what evidence a government team needs to act with confidence.