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
We are running into one issue with this though: downloading the model from github is not so reliable and often only works after a few attempts or times out. Our tool is targeted towards users without much computational expertise, so we automatically download the models. If that occasionally fails it would be quite confusing for users.
Do you think you could host the model weights somewhere else where download is more reliable? E.g. on zenodo? (This is where we host our finetuned segment anything models). If you don't want to do that, would it be ok if I host a copy of your weights somewhere else? (With proper attribution of course.)
Thanks!
The text was updated successfully, but these errors were encountered:
I went ahead and uploaded the model at https://owncloud.gwdg.de/index.php/s/TuDzuwVDHd1ZDnQ since it is not usable for us if hosted on github.
Please let me know if you have any objections and I can remove it (in that case I would encourage you to host it somewhere that enables more stable download).
Hi @ChaoningZhang,
thank you for developing MobileSAM, the work is amazing!
We are developing a tool for interactive segmentation of microscopy images based on Segment Anything here: https://github.com/computational-cell-analytics/micro-sam and would like to add support for your
vit_t
based model. Integration is indeed very easy and I have a PR for it already: computational-cell-analytics/micro-sam#174We are running into one issue with this though: downloading the model from github is not so reliable and often only works after a few attempts or times out. Our tool is targeted towards users without much computational expertise, so we automatically download the models. If that occasionally fails it would be quite confusing for users.
Do you think you could host the model weights somewhere else where download is more reliable? E.g. on zenodo? (This is where we host our finetuned segment anything models). If you don't want to do that, would it be ok if I host a copy of your weights somewhere else? (With proper attribution of course.)
Thanks!
The text was updated successfully, but these errors were encountered: