[FEATURE] using previous frames for inference

[sisse008]

Hello, I have a basic question I was hoping you could help me with.

from the article:

Here, past information can help resolve and constrain the pose estimate. The temporal encoder acts as a generator that, given a sequence of frames I_1, . . . , I_T , outputs the corresponding pose and shape parameters in each frame.

I am new in the field and not very knowledgeable so please bear with me.

my question is: when trying to estimate the pose of object X at frame n, does VIBE use the results(estimations) of X from all the previous frames?

If so, then theoretically, the more previous frames of object X, the better the inference will be at frame n. Am I correct?

Thank you