Does anybody already have (a good) experience with training the new SAM3 model?
I’m getting the following error and I’m wondering if someone else has the same experience or might even already have obtained results? I’d love to see how it performs on my dataset. 
Hi,
I also obtained the same error. Have you managed to find a solution/get help for this error?
Best,
Antoine
Hi Antoine,
Unfortunately I did not. Still waiting. Did you report the bug as well?
Best,
MIlo
Hi Milo,
Yes I also reported the error, but I haven’t heard back since. It’s a little unfortunate, SAM3 seems very promising. Let’s hope that they will be able to solve this error soon.
Antoine
Is this the same on the Chrome browser as well?
Additionally, in my case, I am receiving this error, and I believe it may be because I’m using an uncommon class that is not included in the SAM 3 model’s training data.
I’m working in FireFox, don’t think this is a browser issue. But where are you getting your error @Sagar ? Is this while annotating, training or the new rapid flow?
I am getting error while annotating.
Oke, for me this is new. I have used this feature a couple times and it hasn’t failed yet!
Hi Milo!
Thanks for bringing this issue forward. Our team is working to optimize the SAM3 model training pipeline. For now, I would suggest training SAM3 on 3,000 or less images.
Excellent, I trained on WAY less than that! I started with 50 images or so just to get a feel for it. Are you suggesting that SAM3 is working now?
Thanks!
Hey Milo!
I was playing around a bit with the SAM3 model myself and was able to train on a dataset of a bit under 500 images (segmenting different types of automobiles). Definitely be prepared for a longer training time as this is a very intensive model.
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.
