Halloween may be over, but if you are a 3D printing enthusiast, then you know that does not mean all the monsters are gone. The twisted clumps of filament, known as “spaghetti monsters,” that are produced by a 3D printer when a print goes awry can pop up at any time. Aside from ruined prints, other more serious hazards, like fire, can also arise unexpectedly. As such, it is important to keep an eye on your 3D printer while it is in operation. However, complex prints can take many days to complete, so keeping vigil while your giant Pikachu character prints is not very practical.
As a Hackster News reader, you are probably engineering solutions to this problem in your mind already. In any case, The Spaghetti Detective (TSD) has already beaten you to the punch with a machine learning based solution that cleanly integrates with existing 3D printer control software.
TSD integrates with the popular web-based OctoPrint 3D printer controller and monitoring application as a plugin. It is recommended that you install OctoPrint, and the TSD plugin, on a Raspberry Pi. A webcam is also required to capture images of the print area. After linking OctoPrint to a Spaghetti Detective account, you are all set up for continuous monitoring. A companion phone app can be used to check on the status of prints, and to be alerted about any concerning events.
If you are only printing occasionally, the hosted TSD service is free. For more frequent users it costs a few dollars per month. For the more technically inclined that do not mind putting in a little elbow grease, it is possible to set up a local TSD service, as the entire project is open source. The Raspberry Pi hosting OctoPrint will not quite cut it as a TSD server, however. A NVIDIA Jetson Nano 4GB is suggested, and other platforms with sufficient horsepower can also be used.
The server detects hazards using a convolutional neural network based on the YOLO architecture. These networks are very fast and accurate when it comes to object detection, which is precisely what is needed to detect printing anomalies. And speaking of anomalies, the TSD service has watched over 50 million hours of printing, and has already caught nearly 600,000 failed prints. These stats certainly translate into a lot of saved time and money for TSD users.
If you want an AI to keep an eye on your 3D prints, you can find more information on the Spaghetti Detective website. And if you need to grab the code to set up your own local server, check out this GitHub repo.