Hello fellow reprappers!
For a year or so I've been waiting for someone to develop computer vision algorithms to detect failed prints. Unfortunately I haven't come across any yet. I'm too excited about this idea, so I've decided help out on the development. I truly believe that the thing stopping this from being developed is data. In this case a labeled image dataset of both succeeded prints as well as failed prints.
The full post about this idea and how you can help can be found here:
http://www.fredhedenberg.com/2018/07/02/fdm-3d-printing-image-dataset/
TLDR - Make a dropbox or google drive folder with octoprint timelapses (size less than 50M

. Send a shareable link to
fred@fredhedenberg.com OR share the link here in this thread.
// Fred