Several neural networks, lots of hard work in data labelling, more hard work in data synthesis, even more hard work in dataset curation.. et voila!
Source: it was me with the team who did all of that.
Source: it was me with the team who did all of that.
- extreme number of objects on single photo, typical number of visible pieces in large pile is 1500-2000
- extreme number of classes in multi-class classification, there are ~1000 most common Lego bricks and up to 30000 classes if you include rare bricks and different patterns
- really hard data labelling: one photo can take up to a 5 work days to label
Also you mentioned data synthesis. How would this be possible? Unless your suggesting that you rendered photo realistic piles of Legos and used trained on them because if that is the case, please do a write up of the project. I can't imagine more interesting way to generate training data.
I would estimate that any innovation like this that encourages people to pull out and play with their Lego strongly increases the chances that they'll sell new sets to those people.