How about the one-stage training strategy?

[TheSunWillRise]

Hi, I note that all your models are trained by using two-stage training strategy, which trains models on a larger additional dataset for the first stage. However, in Table 1 from your paper, models you compared with are trained just on DAVIS-2016. THis is not fair for other models, so I wonder how about J mean and F mean of your models trained just on DAVIS-2016?

[seoungwugoh]

Please refer to Table. 5 in our paper for the result without the pre-training.

The DAVIS benchmark does not have a strictly rule on the use of additional training data.
Many previous methods (e.g. MaskTrack, perazzi et al.) adopt pre-training.

Thanks,
Seoung Wug

[TheSunWillRise]

Thanks for your reply. Besides, would you please release your train codes?