How to set up labelbox - issues with Docker

Using Labelbox Annotate
1

I want to analyze/segment my animal behaviour videos for my PhD research with your AI tool:Using Meta's Segment Anything (SAM) model on video with Labelbox's model-assisted labeling. In this tutorial it says that you can just upload the video on Colab notebook, but when I open the website there are no instructions and is really difficult to understand the process. Is there any further tutorial or video or assistance on this? And do I need to have Python installed on my computer?

How fine-scale can segment-anything work? In my videos I have to track the appearance and disappearance of small colour pacthes on the animal while it’s moving, often very small patterns (few pixels). Is this possible to achieve with segment-anything?

I would really appreciate some more guidance in setting up the program and using SAM on my videos…
Thanks a lot!
Gabriella

2

Hi @gabry.scata,

Given SAM2 was released a few days ago we may need to update the guide you have found.
Forgot to mention, you can do that locally or via online tools (Colab, lightning.ai)
What we are doing in the guide goes as follow :

The assumption here are the classes needs to be known hence why we took Yolov8 if you have a specific use cases outside of those classes you would need to train a model for this specific use case.

Many thanks,
PT