Video defect detection · in pilot

Inspection footage, reviewed before anyone presses play.

Upload CCTV, drone or walk-through video and IssuesId’s AI watches all of it. Each defect comes back coded, located along the run, graded and scored for confidence. A person confirms what is real, and confirmed defects become tracked work with the clip attached as evidence. Running now as a pilot with asset owners.

How it works

The review screen on a 45 m, 225 mm cast iron sewer survey. Footage: Wessex Water, “Sewer scanning CCTV footage”, CC BY.

How it works

The AI does the watching. A person does the deciding.

Somebody watching four hours of pipe to find six things wrong with it is the expensive part of a CCTV program. The footage still gets watched, every frame of it, but by the time a person sits down the findings are already in a queue, worst first, each one a click away from the moment it was filmed.

01
Footage in
Upload in the browser, or have the CCTV or survey contractor send it in through the API. What the camera prints on screen (the asset, the start and finish point, the distance counter, the date) is read straight off the video into the survey header.
02
AI proposes
Every frame is reviewed. Each finding comes back with a code from your code list, where it is (chainage, time in the clip, clock position), how bad it is and how sure the model is. A condition that runs along the asset is recorded once, from where it starts to where it stops.
03
A person decides
The queue puts the worst and the least certain findings first, and one click plays the footage from that moment. Confirm, edit or reject, and add anything it missed. Every decision carries the reviewer’s name and the time.
04
It becomes work
Confirmed defects land on the IssuesId register as tracked items or work orders, with the frame and the clip as evidence, issued to the crew that fixes them and closed on proof. The reviewer signs off the condition report.
Industries

Any asset you film, with the codes you already use.

The detection pipeline does not care what the camera was pointed at. What changes from one industry to the next is the footage, the code list and the crew that fixes what it finds. Sewer CCTV coded to the WSA 05 structure is the first code set it is built against; every other row starts from your own code list in a pilot.

SectorFootageWhat it looks forCoded to
Water and wastewater
First code set
Sewer and stormwater CCTV crawler surveysCracks, breaks, deformation, collapse, displaced joints, roots, deposits, infiltration, connectionsWSA 05 Conduit Inspection Reporting Code structure
Roads and civilSurvey-vehicle and dash-mounted videoPotholes, cracking, rutting, worn line marking, damaged signs and barriersThe defect types in your road asset register
Buildings and facilitiesWalk-through video and drone passes of facades and roofsCracking, spalling, water staining, failed sealant, damaged cladding and flashingsThe categories already set up in your IssuesId projects
Construction handoverVideo walks of finished apartments, townhouses and fit-outsDamage, unfinished works and missing items, filed by roomYour snag categories and location tree
Rail, tunnels and culvertsInspection runs through tunnels, culverts and under bridgesSpalling, seepage, cracking, exposed reinforcement, debris and blockagesYour structure inspection manual
Warehouses and logisticsRacking walks and dock-door videoDamaged uprights, bent beams, missing safety locks, floor and dock damageYour racking inspection categories
Energy and utilitiesDrone footage of poles, lines, substations and solar farmsDamaged insulators and cross-arms, vegetation encroachment, broken or soiled panelsYour asset defect codes

Whatever the sector, a confirmed finding lands on the same record the rest of IssuesId runs on: tracked items, work orders, contractors and reports. So a crack found on a facade video and one found on a site walk are closed the same way.

Implementation

Fits how you already work, not the other way round.

Every asset owner arrives with footage in a different place, a different code list, a security team with views on where video lives, and a different appetite for letting software make the call. None of that should decide whether you can use it.

The stamps say what is live in IssuesId today, what runs in a pilot on your footage, and what gets built during one.

Availablelive in IssuesId todayIn pilotruns in a pilot on your footagePlannedbuilt during a pilot, not there yet
  • Hosted in AustraliaAvailable
    IssuesId’s servers and database run in Sydney. For video, footage storage and the AI processing can stay onshore as well.
  • Or inside your own cloudIn pilot
    Where footage has to stay inside your environment, the detection pipeline can be deployed into your organisation’s cloud account on any major provider.
  • Footage from wherever it is nowIn pilot
    Upload in the browser, a contractor sends it straight from their survey software through the API, or a backlog of past surveys comes in as one bulk import.
  • Your codes and your gradesIn pilot
    WSA 05 structure for sewer and stormwater, or the defect list you already use. Condition grades follow the scoring tables you configure, not ours.
  • As much or as little review as you wantIn pilot
    A person confirms every finding, or high-confidence findings in the codes you trust are accepted automatically and a random sample is checked. Set per code family, and changed as the numbers come in.
  • Into the systems you already runAvailable
    A REST API with company-scoped keys and webhooks for events like a defect being raised, available across IssuesId today.
  • Straight onto your network map and asset registerPlanned
    Condition grades and defects pushed to your GIS and asset management system, mapped to your asset IDs. Built in the pilot against the systems you run.
Accuracy

Measured on your footage, before you rely on it.

We do not quote an accuracy figure for your network, because we have not seen your network. A pilot measures it on your own coded footage, code by code, so you decide what to automate from your numbers, not a brochure’s.

01
Send us footage you have already coded
A sample of past surveys, with the codes your reviewers gave them. No new filming.
02
We run it and report the agreement
For each code family: what it found that your coders found, what it missed, and what it called that they did not.
03
You choose what to automate
Full review for the codes that matter most, sampling for the ones it gets right. Your call, made on your numbers.
04
Run it beside your current process
On live surveys, alongside how you review today, until you are happy to switch.
Asked first

Questions asset owners ask about video detection

Does it replace our CCTV coders?
No. It changes their job from watching footage to checking findings. The AI reviews every frame and proposes each observation with a code, its position along the run, a grade and a confidence score; a person confirms, edits or rejects it, and can add anything it missed. Nothing reaches the record without a reviewer’s name on it unless you choose to auto-accept a code family you trust, and even then a random sample is checked.
Which coding standards does it use?
For sewer and stormwater it follows the structure of the WSA 05 Conduit Inspection Reporting Code: the survey header, observation codes with clock position, extent and percentage, continuous defects with start and finish chainage, and structural and service grading. That is the first code set it is built against. For other assets it starts from the code list you already use, and condition grades follow the scoring tables you configure.
How accurate is it?
We do not publish an accuracy figure, because the honest answer depends on your footage, your cameras and your code list. A pilot measures it: we run surveys your reviewers have already coded and report, for each code family, what it found that they found, what it missed and what it called that they did not. You decide what to automate from those numbers.
Where is the footage stored and processed?
IssuesId runs in Australia, and for video the footage storage and the AI processing can stay onshore too. Where footage has to stay inside your own environment, the detection pipeline can be deployed into your organisation’s cloud account on any major provider. Either way your security team gets the detail before a pilot starts.
What footage does it take?
MP4 and other common video formats from CCTV crawlers, survey vehicles, drones and phones, uploaded in the browser, sent by a contractor through the API, or imported as a backlog of past surveys. Where the camera prints the asset, the start and finish point and a distance counter on screen, those are read off the video into the survey header.
What happens to a defect once it is confirmed?
It lands on the same IssuesId register as everything else: a tracked item or a work order with the frame and the clip attached, issued to the crew or contractor who fixes it, chased when it goes overdue and closed on photo evidence by someone who did not do the work. The reviewer signs off the condition report for the asset.
How does a pilot work?
Send us a sample of surveys you have already coded. We run them, report the agreement code by code, and walk you through the review queue on your own assets. If it is worth going further, it runs beside your current review on live surveys until you are happy to switch.
The rest of IssuesId, by industry
Before the next survey program

Send us a survey you have already coded.

We will run it, show you what it found against what your reviewers found, and walk you through the review queue on your own pipes. Then you can decide whether a pilot is worth it.

IssuesId