isi-vista / adam

Abduction to Demonstrate an Articulate Machine
MIT License
11 stars 3 forks source link

Predict diverse objects in scenarios #1179

Closed boyleconnor closed 1 year ago

boyleconnor commented 1 year ago

This is still very messy and incomplete (I haven't touched the training script; the num_objs variable turned out to be unnecessary, will factor out soon) but I wanted to check in sooner rather than later that this is the desired change (of recognizing multiple objects in preprocessing) for later adding spatial relations to ADAM.

EDIT: This PR is now ready for review