This tool should be saving the work being done on labeling incrementally. If your browser crashes (and to test, tried refreshing) you loose all the work you have been doing!
Because of this I will no longer be using roboflow services, as that’s way to much risk on huge datasets.
@WAS Hi, I cannot reproduce. When refreshing my annotations that I have saved are still present. Can you give more information?
This happens when you don’t Save and Continue during the upload process.
The upload process isn’t very clear. It suggests you must have to annotate when you drag and drop or select a folder. And with the number of incomplete datasets (not fully annotated) on the universe I figured this was why.
However, I played around a little more and found you can just click Save and Continue and THEN annotate, where you can refresh, or switch computers, etc.
Ah, I see, thanks for the feedback. We’ll try to streamline this process and add some guardrails here. In the meantime, as you’ve noted, you can feel secure about persisting annotations in the annotate tab
Hi @WAS, I deployed a fix that blocks the creation of annotations at the time of upload (but still allows viewing) to avoid confusion and to make it clearer that you need to save in order to continue the annotation process.
An example of how it turned out: CleanShot 2024-10-02 at 21.32.01 · CleanShot Cloud
I am available for any future questions.
That’s awesome! Great, and quick work here! Thanks so much. Sorry for being so flustered. I had just spent hours working, and then VSCode killed my browser unexpectedly while it was doing stuff.
Kinda random to toss in here, but I noticed there is also no way to reassign images as validation or test? This could be helpful then uploading more images, especially if the ones you intended are already there and refuse to upload.