4 ms·
This is cool. I didn’t know it was possible to use CAD now to build the network(correctly)! If you did not have the CAD files and you were only searching for s
by paulkrush 6y ago
This is cool. I didn’t know it was possible to use CAD now to build the network(correctly)!
If you did not have the CAD files and you were only searching for specific parts you can still build a network that is trained from a single Lego by varying lighting angles and camera pose and spatially correlating that part to itself. If there is huge variation in part types you might also need to infer with known lighting angles and camera angles as well.
I posted a demo on this 4 years ago: https://www.youtube.com/watch?v=mHm4Oz2sml8 https://www.youtube.com/watch?v=mHm4Oz2sml8
BTW: If you know of other references to what I just described I would love to see them!
Over the holidays I built a $10 cable robot to get different cameras poses for this reason:
https://github.com/GemHunt/cable-robot https://github.com/GemHunt/cable-robot
- salamanderman 6y agoHe does say that the simulated images don't really work. He had to supplement the simulated images with 100k manually labelled images. https://towardsdatascience.com/how-i-created-over-100-000-labeled-lego-training-images-ec74191bb4ef https://towardsdatascience.com/how-i-created-over-100-000-la...