4 ms·

Harder by orders of magnitude: https://jacquesmattheij.com/sorting-two-metric-tons-of-lego/ https://jacquesmattheij.com/sorting-two-metric-tons-of-lego/

russellbeattie 8y ago
Wow! 38,000 shapes in 100+ colors (according to that post). Yeah, that's definitely a much bigger challenge. "The final result classifies a part in approximately 30 ms on a GTX1080ti Nvidia GPU." Very nice. I've seen way too many "How It's Made" episodes... He needs some sort of vibrating table, and a high speed belt with a "jump" of some sort to combine with those air cannons.
phaedrus 8y ago
Why I've been meaning to build a "Lego Refractory" instead: a leaf blower at the bottom of a 10ft tall acrylic tube, with exit doors at various heights. It would sort Legos by their terminal velocity in free-fall.
- bigiain 8y ago
  Do it!!! And film every single test run and failure. I would subscribe to this Patreon!!!
- jacquesm 8y ago
  I've tried just that but could not get it to work. If you manage to get it to work reliably this is likely the fastest mechanism you could make. Some problems that you will run in to: - the timing will need to be very precise - you will likely need sensors before the stations - Lego bricks come in all shapes and sizes and will respond unpredictably to air currents - you will need a way to arrest momentum or they'll shoot right back out Still, this is orders of magnitude faster than my contraption if you can get it work.
ataturk 8y ago
My family and I are working on just sorting my son's huge Lego collection into small and medium bins so that we can more quickly rebuild sets and so forth. It's a very hard problem just to come up with a taxonomy that we all agree on! There's not that much info available online, either. It's easy to think about, harder to create understanding among more than one person. Formerly, we had been sorted by color into large bins. Now, we are working on sorting into sub bins by type. We are also separately sub-sorting dark gray and black by type and some by size while other colors can be combined without loss of speed in locating a part. It's hard to explain, but our taxonomy is not as consistent as I would like it to be and it's not clear whether it will help overall or not. This is experimental.

russellbeattie 8y ago

Wow! 38,000 shapes in 100+ colors (according to that post). Yeah, that's definitely a much bigger challenge. "The final result classifies a part in approximately 30 ms on a GTX1080ti Nvidia GPU." Very nice. I've seen way too many "How It's Made" episodes... He needs some sort of vibrating table, and a high speed belt with a "jump" of some sort to combine with those air cannons.

phaedrus 8y ago

Why I've been meaning to build a "Lego Refractory" instead: a leaf blower at the bottom of a 10ft tall acrylic tube, with exit doors at various heights. It would sort Legos by their terminal velocity in free-fall.

bigiain 8y ago
Do it!!! And film every single test run and failure. I would subscribe to this Patreon!!!
jacquesm 8y ago
I've tried just that but could not get it to work. If you manage to get it to work reliably this is likely the fastest mechanism you could make. Some problems that you will run in to: - the timing will need to be very precise - you will likely need sensors before the stations - Lego bricks come in all shapes and sizes and will respond unpredictably to air currents - you will need a way to arrest momentum or they'll shoot right back out Still, this is orders of magnitude faster than my contraption if you can get it work.

ataturk 8y ago

My family and I are working on just sorting my son's huge Lego collection into small and medium bins so that we can more quickly rebuild sets and so forth. It's a very hard problem just to come up with a taxonomy that we all agree on! There's not that much info available online, either. It's easy to think about, harder to create understanding among more than one person. Formerly, we had been sorted by color into large bins. Now, we are working on sorting into sub bins by type. We are also separately sub-sorting dark gray and black by type and some by size while other colors can be combined without loss of speed in locating a part. It's hard to explain, but our taxonomy is not as consistent as I would like it to be and it's not clear whether it will help overall or not. This is experimental.