Volume constrained data about multi_part_assembly

[Wuziyi616]

If I remember correctly, we re-simulated the data with a rejection sampling method, so it is not a subset of the previous non-constrained data.

[eric-zqwang]

Great Thanks.

One more thing I want to clarify. Does it use same data split in the training data and val data for the object between non-constrained and volume constrained?

I mean is there any possible that the non-constrained training data can be found in volume-constrained test data

[Wuziyi616]

I believe so, the data split should be the same. You can check with the split file to see if there is overlap in the object ID.