Computer Vision Deployment
The course on getting computer vision to work on a real site — not just in a demo.
A practical course for the people who own the deployment: from the first feasibility check to a network of sites. Built on True Detection’s international deployments.
01 — The gap
The demo always works. The site doesn’t.
Between a lab metric and production sits a gap that costs companies pilots, budgets and trust in the technology.
- Rain, backlighting, partial occlusion, an unusual posture.
- Cameras mounted at the wrong angle.
- Alerts nobody acts on.
This is not a course about computer vision as a technology. It is a course about deployment. The structure follows the project lifecycle, and theory is covered only to the depth you need in order to make decisions. Cases are not illustrations at the end of a module — they are the load-bearing structure: every module ends at a decision point drawn from field practice.
02 — Audience
Who it is for
Built for the people who own the purchase and the launch — decision makers and the project leads who report to them. No technical background is required. In Module 0 you take a short entry diagnostic and place your company; your answer determines which modules matter most to you.
-
01
No cameras yet
Start with feasibility, economics and site readiness before you buy.
-
02
CCTV already in place
Learn what can be reused, what has to be replaced, and why.
-
03
A pilot running
Write success criteria, a baseline and acceptance before scope creeps.
-
04
Production deployment
Plan operations, people, privacy and the path from one site to a network.
03 — Output
You leave with a folder of working documents, not lecture notes
Every module closes with an artefact you can put to use on your own site the same day.
-
01
Feasibility Screening Card
A one-pager for testing an idea for viability.
-
02
ROI Calculator (Excel)
With preset scenarios for three industries.
-
03
Site Readiness Checklist
A 40–50 point field checklist you walk your own site with.
-
04
Data Collection Plan
A template that costs out data volume, timelines and annotation budget.
-
05
Pilot Charter
Two pages: scope, metrics, baseline, acceptance criteria, roles.
-
06
Response SOP Template
And an escalation map — who acts when an alert fires.
-
07
Stakeholder Map
And a deployment communication plan.
-
08
Privacy Impact Pre-Screen
A 15-question screener before you involve counsel.
-
09
Scale-Up Roadmap
A template across three horizons.
04 — Curriculum
Program: 11 modules and a capstone
Follow the complete deployment journey — from assessing feasibility and ROI to piloting, integration and scale. Each module turns a critical decision into a practical tool you can apply to your own site.
-
Module 0
Where you are now, and where you will be by the end. An entry diagnostic places your company on a maturity scale and sets your personal route through the course.
Artefact: Entry Diagnostic Questionnaire.
-
Module 1
Realistic expectations instead of pitch-deck promises: a map of what you can actually ask for, what each kind of task costs in effort, and what breaks it on a live site. And why an MVP that satisfied everyone in the room behaves differently once it is running.
Artefact: Feasibility Screening Card.
-
Module 2
Run the economics before you buy, not after — where the value actually comes from, and which cost lines customers consistently underestimate. The core idea of the module: computer vision does not create value, it creates a signal. Value comes from the changed response process.
Artefact: ROI Calculator (Excel).
-
Module 3
A diagnosis of whether your site is physically ready: cameras and angles, lighting, network, and where the processing should run. What can be reused from existing CCTV and what has to be replaced — usually more than anyone expects, and far cheaper to find out now than during a pilot.
Artefact: Site Readiness Checklist.
-
Module 4
Why a model trained elsewhere underperforms on your site, how much data you actually need, and what annotation really costs in money and time. Including the contractual question of who owns the data and the trained model — settled before the pilot, not after.
Artefact: Data Collection Plan.
-
Module 5
Success criteria get written before the start. How to measure a baseline, which metrics let you talk to a vendor as an equal, how long a pilot has to run to mean anything, and what belongs in the acceptance clause.
Artefact: Pilot Charter.
-
Module 6
How an alert becomes an action: where the signal goes, who responds, within what time, and what gets logged. Alarm fatigue, and why a sensitivity threshold set by engineers rather than by operations quietly kills the system.
Artefact: Response SOP Template and escalation map.
-
Module 7
Resistance rarely comes from the shop floor. In documented deployments it comes from the supervisory layer — the people whose judgement the system starts to audit. How to position the system, train operators and set supervisor KPIs so the project survives its first quarter.
Artefact: Stakeholder Map and communication plan.
-
Module 8
Where the line between anonymised detection and biometric identification actually falls, and why a vendor’s assurance that a system “is not biometrics” is a commercial statement rather than a legal classification. An orientation across the regimes you are likely to operate under — not a legal opinion.
Artefact: Privacy Impact Pre-Screen.
-
Module 9
What to do once the pilot has worked: why accuracy degrades over time, what a retraining loop costs, and how to keep configurations consistent across sites. Total cost of ownership over three to five years, and the governance question of who owns the platform inside the company.
Artefact: Scale-Up Roadmap across three horizons.
-
Module 10
A full module, not an appendix at the back. Six post-mortems, each traced to a different root cause and mapped back to the module that would have prevented it — in the same format every time: what was planned, what happened, why, and what should have been done instead.
Artefact: Problem Tree and Risk Register.
-
Capstone
You defend a deployment plan for a real site at your own company, assembled from the work done through the course. Ten minutes, with feedback from the instructor and the group, and a document ready for internal sign-off.
Artefact: A full deployment plan for a real site in your own organization.
05 — Format
How the course is built
-
11 modules and a capstone
The project lifecycle in sequence, from orientation to defence.
-
Delivered in English
The course language for every module, case and capstone defence.
-
9 working artefacts
Checklists, templates and an ROI calculator you can use on site.
-
Capstone defence
A ten-minute presentation with feedback from the instructor and the group.
-
Certificate on completion
Issued after the capstone defence — for teams whose training budget sits separately from the project budget.
-
What you walk away with
A deployment plan for a real site
Assembled from the course artefacts and ready for internal sign-off at your company.
06 — Questions
Before you apply
No. It is a course about deployment. Theory is covered only to the depth you need in order to make decisions. The structure follows the project lifecycle, from feasibility to a network of sites.
No. The course is built for the people who decide and the people who run the project, not for engineers. You will learn what pixel density means because you cannot judge a camera proposal without it. You will not learn how to train a model, because you are not going to train one.
No. The course is built to work at any level of maturity: no cameras yet, CCTV already in place, a pilot running, or production deployment. Module 0 sets your personal route.
Nine working artefacts you can use on your own site the same day, and a capstone deployment plan assembled from those artefacts, defended in a ten-minute presentation.
No. Module 8 gives orientation, not a legal opinion. Regulation changes, and the final check is done by local counsel.
English.
No. Grant participants sit in the same cohort, do the same artefacts and defend the same capstone. We do not tell the group who is on a grant.
One seat per organisation per cohort. If the deployment needs several people trained, that is a conversation about the paid program — write to welcome@aiplomacy.com.
07 — Apply
Enrol now
Tell us who you are and where your deployment stands. We use this form to follow up on the program — not to send a pitch deck.
08 — Grant
Grant: a funded seat on the program
Who can apply
-
01
A site you can name
You are running, or about to start, a deployment at a specific site. Not a general interest in the technology.
-
02
You own the project
You decide on the purchase or you run the launch. The grant seat is not a sampler for someone collecting information.
-
03
No training budget this cycle
Your organisation cannot fund the seat. We ask for no financial documents and verify nothing — we read what you write.
The grant covers the program fee in full: all 11 modules, the nine artefacts, the capstone defence with feedback, and the certificate. One seat per organisation per cohort.
We read every application. What decides is how specifically you describe your site and what you intend to do after the capstone, not your job title or your company’s name. Decisions go out 5 working days after application — everyone receives an answer, including the ones we cannot fund.