The value of the dynamic_ratio in GMM

[xiyuan02]

I have noticed that the file in cfg sets the value of the dynamic_ratio to 1，I want to know whether it is effective？ for i, proposals in enumerate(proposal_list): dynamic_ratio = self.train_cfg.dynamic_ratio scores = proposals[:, 4].clone() scores = scores.sort(descending=True)[0] if len(scores) == 0: thrs.append(1) # no kept pseudo boxes else: # num_gt = int(scores.sum() + 0.5) num_gt = int(scores.sum() * dynamic_ratio + 0.5) num_gt = min(num_gt, len(scores) - 1) thrs.append(scores[num_gt] - 1e-5)

[Johnson-Wang]

Thanks for your interest. The initial code only set num_gt = int(scores.sum() + 0.5) as default. After submitting the paper, we also tried to add another dynamic_ratio to control the number of gts from each image. We found that the default value 1 worked best and thus stopped there. Yet personally, I think this value could also be tuned on other datasets as well.

[xiyuan02]

Thank you very much, and I have an other question: how do you confirm the value of N in GMM?