Hi @ShoufaChen , thanks for sharing the great project.
From the paper, it's said that box renewal operation is used as
Directly sending these undesired boxes to the next sampling iteration would not bring a benefit since their distribution is not constructed by box corruption in training.
However, the paper also said only one sampling step is applied during inference.
Dose this mean the box renewal actually has no effect on the inference procedure?
Hi @ShoufaChen , thanks for sharing the great project.
From the paper, it's said that box renewal operation is used as
Directly sending these undesired boxes to the next sampling iteration would not bring a benefit since their distribution is not constructed by box corruption in training.However, the paper also said only one sampling step is applied during inference.
Dose this mean the box renewal actually has no effect on the inference procedure?
Thank you.