You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have an issue when reproducing results on CityScapes with released checkpoints, metrics are worse than in paper.
Can you provide infromation how to achieve results given in paper? I noticed some incosistency between CityScapes checkpoints names regarding count of training epochs specified in paper, maybe this is the issue and you can provide another checkpoints.
Thank you!
The text was updated successfully, but these errors were encountered:
Hi, thanks for your interest in our work! Please check this training docs. We present some different choices when we reproduced the experiments and we also recommand people to report these numbers in follow-up works.
Hello, thank you for the interesting work!
I have an issue when reproducing results on CityScapes with released checkpoints, metrics are worse than in paper.
Can you provide infromation how to achieve results given in paper? I noticed some incosistency between CityScapes checkpoints names regarding count of training epochs specified in paper, maybe this is the issue and you can provide another checkpoints.
Thank you!
The text was updated successfully, but these errors were encountered: