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
Hello again !
Not really an issue with the current version, but more a curiosity about the future improvements :
Have you considered using the Fast-SAM model to decrease the encoding time for large geospatial images ? According to the developer's team, it seems to have comparable performance at 50× higher run-time speed compared to the original SAM model.
Here is the Github repo : https://github.com/opengeos/FastSAM
Thanks
The text was updated successfully, but these errors were encountered:
Hi @4del-Yousefi@cycle4 . Thank you for your suggestions. We have actually been paying close attention to these new SAM-based projects for a while: FastSAM, HQ-SAM, MobileSAM. But considering these projects are trained on much smaller datasets compared to the original SAM, their generalization ability may not be as strong as SAM's. Also, since we are not full-time developers and have limited time for development, we have not tried these models yet. However, they are in our plans. We are especially interested in the HQ-SAM project. It seems to address some of SAM's shortcomings to some extent, and has also integrated the work from MobileSAM by releasing Light HQ-SAM that can encode images faster. So we have added it to our Future Works.
Hello again !
Not really an issue with the current version, but more a curiosity about the future improvements :
Have you considered using the Fast-SAM model to decrease the encoding time for large geospatial images ? According to the developer's team, it seems to have comparable performance at 50× higher run-time speed compared to the original SAM model.
Here is the Github repo : https://github.com/opengeos/FastSAM
Thanks
The text was updated successfully, but these errors were encountered: