Getting everything set up was way more complicated than we thought it'd be, but we kept track of the instructions well enough that Team MIT was able to replicate our results! (Well, that and I suppose we have very capable collaborators :))
Initial results. Again, this is the baseline (i.e., focus is on getting infrastructure set up, not robustness or accuracy), so what we'll have in the future will be so crazy good that no one else should enter the competition. The red blob of dots with the blue arrows sticking out is the predicted location of the object along with its orientation.
This result looks pretty good.
We're just about there. At least it didn't confuse our teammate's arm with the box...
Hey, not bad! It's a match!
Not bad, but not great either. We don't really see all that much of the green box in the image.
Ok, so this prediction isn't that great.