Skip to content
OffsightAI
Sign In

Vision AI

Your cameras,
turned into live factory data.

Connect cameras in minutes, train Vision AI on your own floor, and see machine states, production progress, tools and work-area activity flow straight into Offsight, ready to act on.

  • Cameras connected in minutes
  • Trained without a data team
  • Console, iOS and Android
A fixed camera’s view of a wiring-assembly work area, highlighted in blue and marked active, with a switchgear cabinet outlined in green as the product and a tool cart outlined in orange as equipment
  • Machines
  • Products
  • Work areas
  • Tools & equipment

The setup, step by step

From a camera on the wall to live factory data.

Five steps, most of them measured in minutes. Click through the flow, screen by screen, in the Offsight console and mobile apps, or press play.

Step 1 · Internet of Things

Add a camera in Internet of Things

Connect cameras · screen 1 of 3

1 / 15
The Offsight Internet of Things page with the guided camera setup open, one step per stage from naming the camera to mounting it

Internet of Things in the Offsight console, with a camera setup open.

Choose Add Device, pick the camera, and Offsight opens a guided setup for that exact model.

No integrator, no network project

Power the camera, pair it from a phone, join the plant Wi-Fi and paste Offsight’s upload details into the Reolink app. Frames then arrive on their own.

Internet of Things on iPhone: a connected Reolink camera sending a fresh image every 30 seconds

A connected camera on iPhone, sending a still every 30 seconds.

Minutes

To connect a camera in Internet of Things, every setup step saved as you go

3 frames

Marked per camera to set up a work area for training

12 + 12

Marked On and Off frames to train a machine’s states

3 apps

Console, iOS and Android, from camera setup to the live feed

What it understands

Machines, products, work areas, tools and equipment.

One camera sees all of it. Vision AI turns each into data Offsight can act on.

A machine floor with an enclosed laser cutter, a press brake and a CNC machining center

Machines

Every change of state, recorded against the machine record as it happens.

  • On and Off
  • In Use
  • Under Repair
  • Your custom states
A final-quality area with a switchgear cabinet highlighted in green as the product in progress

Products

The unit on the floor and the work area it is in, so progress shows up without a scan.

  • Unit in progress
  • Work area entered
  • Time in each area
  • Hand-offs
An engine-integration work area highlighted in blue around a generator outlined in green, with a tool cart outlined in orange

Work areas

Draw the area once. Vision AI reads the activity inside it, frame after frame.

  • Active or idle
  • Work in progress
  • Tools present
  • Labor activity
A final-inspection area with a tool cart and drill outlined in orange beside the inspection bench

Tools & equipment

Carts, welders, grinders and portable equipment, picked out in every work area’s frames.

  • Tool carts
  • Welders
  • Grinders and drills
  • Test equipment

See it the way Vision AI does

Same camera. A completely different picture.

Drag across the frame. On the left, the raw still. On the right, the unit in progress that Vision AI picks out of it.

A raw camera still of a functional-testing area with a switchgear cabinet, test benches and a cable reel
The same still as Vision AI reads it, with the switchgear cabinet highlighted in green as the unit in progress

Machine states

Know which machines are running, without asking.

Vision AI watches each trained machine and records every change of state the moment it is sure.
  • On and Off to start, then In Use, Under Repair, Needs Repair or any custom state your team uses.
  • A confidence floor and a run of agreeing frames stop one odd picture from logging a false change.
  • Every change is written to the machine record, building uptime and downtime history on its own.
  • Review and improve the training any time, from the console or the phone.

See the machine register

Machine states · today6 AM – 6 PM
Laser Cutter 0384% on
CNC Router 0255% on
Press Brake 0154% on
  • On
  • Off
  • Under Repair
  • Changeover (custom)
An illustrative shift, built from the state changes Vision AI records.

Production progress and equipment

Follow the work across the floor.

Each work area’s camera records what is in it, frame after frame, so you can see when a unit or a cart was in each area.

A switchgear lineup, from wiring to final quality

Three work areas, three cameras. Each frame shows which area the unit is in.

  1. A switchgear cabinet outlined in green in the wiring-assembly area

    Wiring Assembly7:42 AM

  2. The switchgear cabinet outlined in green in the functional-testing area

    Functional Testing11:18 AM

  3. The switchgear cabinet outlined in green in the final-quality area

    Final Quality2:36 PM

The drill cart, wherever the work takes it

See which work area the shared cart is in, without searching the building.

  1. A tool cart and drill outlined in orange in the electrical-assembly area

    Electrical Assembly8:05 AM

  2. The same tool cart outlined in orange in the mechanical-assembly area

    Mechanical Assembly10:52 AM

  3. The tool cart outlined in orange at the final-inspection bench

    Final Inspection3:14 PM

Instantly in Offsight

What the cameras see becomes what your team does next.

Vision AI is not another screen to watch. Its signals land in the same Offsight system that runs production, quality, materials and labor.

Real-time actionability

A machine goes down, a unit stalls, a work area goes quiet. The right person hears about it while it still matters, on the console or the phone.

Operational management

Supervisors re-sequence work, move people and send technicians from live floor data instead of a walk-around.

KPI and production analytics

Uptime, time in each work area and throughput come from what the cameras saw, not from end-of-shift data entry.

Deployment

Cameras on the wall. Everything else is Offsight’s job.

No servers to rack, no GPUs to buy and no machine-learning pipeline to maintain.

Off-the-shelf cameras

Supported Reolink cameras mount in minutes. A dual-lens model covers a 180° view, so one mount can watch a whole bay.

Nothing to host on site

Cameras upload stills to Offsight over encrypted FTPS. Offsight trains the models and runs them for you.

Trained on your floor

Each model learns from your own frames and the names your team already uses, so it knows your machines, your products and your work areas.

For the whole team

Set up, mark up, review and act from the console, the iPhone or the Android app, with the same data everywhere.

Where to buy?

Two cameras. Both ready for Vision AI.

Vision AI works with the two Reolink cameras Offsight sets up in Internet of Things. Pick one up from Reolink, Amazon or Walmart, and it is watching your floor the same day.

Questions

Vision AI, answered.

How long does it take to connect a camera?

Minutes. Internet of Things walks you through naming the camera, powering it, pairing it over Bluetooth, joining the plant Wi-Fi, entering Offsight’s FTPS upload details, binding it and verifying the first frame. Every step is saved, so you can stop and pick up later.

Do we need a data-science team to train it?

No. Your team marks up a handful of real frames (three per camera for a work area, twelve each for a machine’s On and Off states) and presses Start training. Offsight trains the model on its own GPUs and deploys it for you.

Does Vision AI stream video?

No. Cameras send stills through the shift. That keeps the load on the plant network light and gives every moment a clean, timestamped frame you can scrub back to.

What can it recognize?

Machine states such as On, Off, In Use and Under Repair, plus custom states you define; the products in each work area; tools and portable equipment; and activity inside the work areas you draw.

Which devices can we use?

Camera setup, the live feed, markup, training and review all work in the Offsight console and in the iOS and Android apps.

Do we need servers or GPUs on site?

No. The only hardware on site is the cameras. Offsight trains and runs the models, and the results land directly in your Offsight account.

Factory camera scenes on this page are illustrations, including the frames shown inside the camera-feed and markup screens. The screens themselves are from the Offsight console and iPhone app.

See Vision AI on your factory floor.

Bring a camera angle, a floor plan or the machine you most want to watch. We will show you the setup, the live feed and the model, end to end.