Robot arm experiment needed

Wondering if anyone has an idea for a good “beginner” experiment for integrating deep learning, computer vision and a consumer 4-DOF robot arm (programmable in Python)? I’m still pretty green in all these areas - especially the robotics, and a lot of the relevant papers I have read so far, are definitely far above my skill level and/or require a huge data set that wouldn’t be easy to recreate. But I would like something that I can use to progress in all these areas at once. Obviously needs to be nothing trailblazing.

I was thinking maybe something like recognizing board game pieces and picking them up might be fun. But maybe even that will prove too difficult. Anyone have any pointers/ideas?


1 Like


A long time ago we built Roboartist:

Roboartist used something of a modified Canny edge detector implemented in Matlab. I now think its time for a version 3, but haven’t seen any really good papers on photo to line drawing using deep learning. I’m definitely interested in this area, so let me know if this is the kind of thing you’re looking for.

You can read more about the project here:


Thanks, I will read up on it! It looks like a very cool project!