A computer vision platform that turns recorded and live video feeds into people, vehicle, and maritime operational insights.
Interface studyA multi-use-case computer vision system with operational trend views and alerting workflows.
Overview
A useful system starts with a clear reason to exist.
01 / What was built
A computer vision platform for people, vehicle, and maritime operational insights.
02 / Who it was for
Operations teams with camera infrastructure but limited decision visibility.
03 / Why it mattered
Recorded and live footage could become trends, alerts, and usable operational evidence.
Business challenge
The question was not what technology to use. It was what needed to work better.
Organizations had camera infrastructure but limited ways to turn footage into useful decisions, monitoring trends, or timely operational alerts.
Plumfind role / research, strategy, design, engineering, deployment, and support planning.
Thousands of cameras. Millions of hours.Zero operational insight.
Discovery & research
Reduce uncertainty before production.
01
Business context
Organizations had camera infrastructure but limited ways to turn footage into useful decisions, monitoring trends, or timely operational alerts.
02
Research
We identified reusable use cases across transportation, logistics, smart infrastructure, retail, and facilities before defining the shared platform foundation.
03
Experience direction
The experience focuses on the decision layer: clear counts, trends, modes, and alerts rather than an interface that simply exposes detections.
04
Production plan
It was deployed on AWS using EC2, an application load balancer, ACM, Route 53, and secure hosting controls.
Research artefact
Clarity lives in the details.
We identified reusable use cases across transportation, logistics, smart infrastructure, retail, and facilities before defining the shared platform foundation.
PORT OF MONTREAL / LIVEVessel detectedAlert routed to operations
Activity over time
Solution
What Plumfind designed and built.
The experience focuses on the decision layer: clear counts, trends, modes, and alerts rather than an interface that simply exposes detections.
01
Reusable detection foundation
A shared platform was designed to support multiple industry use cases rather than a single narrowly scoped deployment.
02
Decision-first analytics
Counts, trends, monitoring modes, and alerts were prioritized over a raw stream of model detections.
03
Maritime specialization
Custom fine-tuning extended the platform to a business-specific vessel detection workflow.
Architecture
The system is a set of deliberate connections.
Each layer was selected to support the business workflow, not to make the technical picture more complicated.
01Camera feeds
02Detection and tracking
03Analytics layer
04Secure cloud delivery
01
Camera feeds
Recorded and live video provide the operational source material.
02
Detection and tracking
YOLO and tracking models identify movement, objects, and relevant events.
03
Analytics layer
Events become counts, trends, dashboards, and routing logic.
04
Secure cloud delivery
AWS hosting, load balancing, certificates, DNS, and controls make the system available in production.
Implementation
From considered direction to a production system.
01
Experience and system design
The experience focuses on the decision layer: clear counts, trends, modes, and alerts rather than an interface that simply exposes detections.
02
Engineering
The platform combines YOLO-based object detection, tracking, custom maritime fine-tuning, analytics views, and live-stream processing.
03
Deployment and ownership
It was deployed on AWS using EC2, an application load balancer, ACM, Route 53, and secure hosting controls.
Business impact
What changed because the system became clearer.
01
Delivered people, vehicle, and vessel detection workflows
02
Created operational trend dashboards
03
Implemented custom maritime detection
04
Connected live monitoring to automated alerts
Lessons learned
Detection alone is not operational intelligence
A shared platform can support multiple industries
The output needs to answer a business question
Technologies used
Selected for the work they needed to do.
Technology choices were made around fit, maintainability, and the operating reality of the project.
01
Vision intelligence
YOLO, object tracking, Roboflow
Selected for adaptable object detection, tracking, model operations, and specialized fine-tuning.
