QuinlyVision AI Failure Detection
QuinlyVision is a machine learning algorithm that we trained to look for 3D printing defects and failures.
When an issue is detected, you'll get an email alert so you can check and stop the print.
In future versions, QuinlyVision will be able to automatically pause your print for you, and eventually may even be able to automatically correct the issue in real-time, saving filament and avoiding a huge mess.
QuinlyVision detects and responds to failed prints, but it is not a replacement for a fire-safety system. Be smart. Take appropriate precautions, and don't leave 3D printers unattended.
QuinlyVision requires no plugins or additional installation.
QuinlyVision runs 24/7, and is a free service. You can use QuinlyVision 24/7, 365 days of the year, for all of your 3D printers, at no extra cost!
Currently, 5 failure types are detected:
- Failed first layer
More failure types will be added as they improve.
The Future of QuinlyVision
- More informative error notification emails
- Better ways to provide feedback for missed fails & false positives
- Automatic pause when failure is detected with high confidence
- Toggle for notifications & auto-pause for each failure type
- Machine-learning model accuracy improvements
- UI improvements
- All 14 failure types
- Custom responses depending on failure type (adjust temps, speed, acceleration, etc.)
- Option to automatically clear bed and retry print, or move on to next item in queue.
- Automatically suggest and make changes to gcode to avoid reslicing.
Start Using QuinlyVision
All you need to do is plug in a webcam to your Pi / Hub! QuinlyVision is enabled by default, and will run on all of your prints unless you disable it.
To see what QuinlyVision sees, click "Open AI View" and there will be a red box around anything detected as an issue.
Use the sensitivity slider to adjust how easy it is to trigger an alert.
- 10 is very high sensitivity, it will trigger more often, and very quickly after an issue appears, but there will be more false-positives.
- 1 is very low sensitivity, it will trigger less often, and might miss obvious failures sometimes.
- Usually, it's best to stay somewhere in the middle, although you should experiment to find the setting that works best with your setup.
Machine learning algorithms cannot be 100% accurate or reliable. You might occasionally get an alert when your print is fine. This is called a 'false-positive' and it means QuinlyVision thought your print has failed when it hasn't.
If your 3D printer looks like this, then you can't really blame QuinlyVision for getting it wrong...
There are ways to reduce the risk of a false positive. Set QuinlyVision up for success by following the recommendations in the Webcam Setup Guide