Edge quality checks for factory lines. Packaged modules decide pass or fail next to your cameras — labels, caps, safety.
Line 3 — Bottling
--:--
Inspected
48,201
Pass rate
99.2%
Defects
386
Defect flagged on Line 3
4 open
Label misaligned — 7.4° skew
Cap missing
Short count 11/12
Fill level below spec
edge-01 · healthy
38 ms p95
IFA QA (In-Factory Analysis — Quality Assurance) is an edge computer vision platform for production lines. Each check ships as a module — one task on one line, such as label misalignment or missing caps on a bottling line, or a person entering a guarded zone.
A module runs on a small edge box beside your existing cameras. It decides pass or fail on site and emits a standardized event, so the line keeps running whether or not the network does.
Events stream to a cloud dashboard: live line and camera health, analytics derived automatically from each module, and a full audit trail of every check.No-code automations turn events into action — filter, window, and cooldown rules that trigger email alerts, frame captures, or reports, without writing code or touching the box.
Each line check ships as a packaged module — label misalignment, missing caps, person and safety detection. Adding one is configuration, not a new project.
Inference runs on an edge box beside your cameras. The verdict leaves as an event; the video stays on site.
Connect any event to an action — email alerts, frame capture, reports — with every run logged step by step.
The IFA dashboard
IFA QA and IFA Counting report into the same web dashboard. It is built so a new module shows up with its charts, events, and automations the day it is deployed — nothing is coded per module.
Plant overview — today
Events per hour
Events by severity
184 events · 2 modules · 4 cameras
Streams
Automation runs
Every camera as a health card: online, degraded, or offline, with frame rate, latency, and uptime, and a live view from the edge box when it is streaming. Refreshes on its own.
Pick a metric, group by line, zone, severity, or product, and bucket by hour, day, or week. Line, bar, area, stacked, and pie — all built from the events a module emits, never hand-coded.
Every event, filterable by module, type, and severity, with the frame that triggered it. Email and WhatsApp notifications go out through automations, with a per-run record of what was sent and what was suppressed.
A drag-and-connect flow builder: a trigger (a module event, a camera going offline, a schedule), logic (filter, window, cooldown, time gate), then actions — send an email, send a WhatsApp message, capture the frame, generate a report.
A report action turns a session or a shift into a readable document — totals, shares, per-product detail — attached to the notification email and kept in the dashboard's report list.
IFA QA questions
Modules in production cover label misalignment, missing caps, and person detection for guarded zones. Each module targets one task on one line and returns a pass or fail verdict with the frame that decided it. New modules are built around what your plant needs — and quantity counting is its own product, IFA Counting.
Your existing cameras and a small edge box beside the line. The box takes a camera index, a video file, or a stream URL as its source, then asks the cloud which module it should run and loads it at runtime — so changing what a line checks does not mean a site visit.
You build a flow in the dashboard: a trigger (a module event, a stream going offline, a schedule), then logic (filter, a window such as 'five or more in ten minutes', a cooldown), then an action (send an email, capture the frame, generate a report). Every run is recorded with per-step detail, including runs that were suppressed.
No. The live view is an in-memory latest-frame stream that is never written to disk, and the events sent to the cloud carry data, not media. Frames are only stored when you explicitly add a frame-capture action to an automation.
Last updated:
