I wonder why there is a different result between 0-shot, 10-shot and 25-shot when use synthetic data(Table2 (a) rows 2 and 3). I think the differnt setting of YCBVtrain has no influence over the synthetic data generation and training progress of detector. Or does the (pre) means that detection model is pre-trained on each few-shot YCBVtrain?
I wonder why there is a different result between 0-shot, 10-shot and 25-shot when use synthetic data(Table2 (a) rows 2 and 3). I think the differnt setting of YCBVtrain has no influence over the synthetic data generation and training progress of detector. Or does the (pre) means that detection model is pre-trained on each few-shot YCBVtrain?