AI Camera Assist
Attendance, behavioural analysis and task completion, from ordinary camera footage.
–
Clips ingested?Footage the engine holds, sampled into frames and ready to analyse.
0
Analysed by model?Clips a model has labelled window by window and the grammar has ruled on.
–
Task workflows?Task definitions: the steps and the order rules between them.
0
Steps defined?Every named step across every workflow, e.g. wash, cut, fold in half.
New here?
Start on Sources to bring footage in, then Workflows to describe the task. How it works explains the whole system from scratch if any of this is unfamiliar.
How a verdict is produced
The model is asked one narrow question per window of video: which step is this? Whether the observed order is acceptable is decided separately, by a rule set you can read off a page. That split is deliberate. Order rules are the part a care supervisor will challenge, and they have to give the same answer every time.
1
Ingest
Camera clip is sampled into evenly spaced frames
2
Classify
Model labels each window with the step it shows
3
Verify
Grammar checks the order against the workflow
4
Review
A person accepts, corrects or rejects the verdict