Visual inspection software you can test before you buy hardware.
AI defect detection on your own parts. Upload images and video of good and defective parts, label the defects with one click or by asking Vero, and train a detector. Test it on new parts before anyone relies on it.
What AI visual inspection changes.
Catch defects before they ship
Train on your own parts and prove the model on new ones before it decides anything.
Steadier quality
A model applies the same check to every image, on every shift.
Decisions with evidence
Every prediction shows its box or outline, its class and its confidence.
Automated visual inspection software, step by step.
What the defect detection software gives you.
A trained model
Classification, detection or segmentation
Evidence
Accuracy per model and test predictions
An API
Run Model, Run Inference, Stop Model
A versioned dataset
The spec for your integrator
AI defect detection, instead of inspectors, rules and integrator quotes.
Pick your role.
Quality engineers
Stop escapes without over-rejecting.
- Detect and locate defects on your own parts
- Model accuracy, and Test on held-out parts
- Evidence on every prediction: outline, class, confidence
Automation & vision engineers
Rules keep breaking on variation.
- Train on examples, not thresholds
- Mini (CPU) or Plus (production) models
- Call the model from a test script over REST
Plant & quality managers
Proof before capex.
- Try it on your own parts for $0
- Pro at $20/mo adds the API
- A named path to production with Averroes
When it fits, and when it doesn’t.
A good fit
- Feasibility before buying cameras or integrator time
- Offline analysis of rejects and escapes
- Building the dataset and model for the line
- Batch or near-line checks over the API
Use something else when
- Real-time pass/fail at line speed (Averroes production inspection)
- PLC or MES integration out of the box
- Air-gapped or validated GxP lines
Visual inspection AI, answered.
Our images can't leave the plant.
VisionRepo runs in the cloud. Use it for feasibility and offline data; inline, on-prem and air-gapped inspection is Averroes production inspection.
We don't have defect images.
Anomaly detection trains on images of good parts only; under 100 images it trains a few-shot model. It's on the Ultra plan, because it trains with the Extensive time budget and the Plus size.
How do we validate a model?
Each model shows its accuracy, and Test runs it on held-out parts with every prediction and its confidence. Formal evaluation sets and monitoring on the line are part of Averroes production inspection.
Can it run at line speed?
Not from VisionRepo's cloud. Prove and build the model here; line-speed decisions run at the edge with Averroes.
Is VisionRepo AI inspection software for the line?
It is AI inspection software for building and proving the model: manufacturing defect detection on your own images, tested before anyone relies on it. Line-speed decisions run with Averroes production inspection.
Anomaly detection learns from good parts.
Upload images of good parts only. The model flags anything that looks different. Under 100 images it trains a few-shot model.
Good part
FlaggedPut your visual data to work.
Free plan, no credit card. Bring your own images or start from a sample.