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
Many thanks for releasing the codes.
We try to implement the results of SatMAE on the fMoW-Temporal dataset using default setups on your paper. However, we only achieve performance around 74.453/91.314. Are there any clues regarding how to achieve comparable evaluation results compared to the reported results (81.49/93.26)?
The text was updated successfully, but these errors were encountered:
we have noticed that a few of the training images (less than 10) in the provided metadata files(.csv) are not readable on the current fMoW-Temporal dataset. This could be one of the reasons for our implementation but seemingly not the major issue.
Many thanks for releasing the codes.
We try to implement the results of SatMAE on the fMoW-Temporal dataset using default setups on your paper. However, we only achieve performance around 74.453/91.314. Are there any clues regarding how to achieve comparable evaluation results compared to the reported results (81.49/93.26)?
The text was updated successfully, but these errors were encountered: