Open zhaoguangxiang opened 4 years ago
Learn object representations and their relationship is intersting, but it is hard to understand how can you know all objcets at the begining for all examples.
To my knowledge, mostly in such object representation learning works, what you actually set as a hyperparameter is an upper bound of the number of objects that may appear in the scene at any time. And the network will learn how to represent an 'empty' object, which is a vector representation with no influence on the reconstruction, interaction or whatever.
Learn object representations and their relationship is intersting, but it is hard to understand how can you know all objcets at the begining for all examples.