In the “Catalog” section, I have a Dataset named “Google” with screenshots of websites that feature the Google logo. However, this dataset also contains screenshots that are unrelated to the Google brand, and I need to separate them and create “Slices”. How should I correctly do this, considering that my dataset with Google screenshots is constantly being updated with new images, and there will be more irrelevant pictures that need to be added to the existing “Slices”?
Additionally, on the attached screenshot, I would like to know what these values represent and how to set them. In other words, these “Slices” should filter out screenshots that are unrelated to the Google brand.
If anyone can help me with this?
Thanks!
Hey @mrevina22 ,
Is your Dataset (“Google”) receiving any metadata with the image files when they are uploaded? If not, then you will need to build a step to do that in your Dataset.
If you define your important slices as:
-Google
-Not Google
Then you could achieve that by creating an Ontology including (at least) a global classification feature in Schema and then send a sample of your Dataset to an Annotate Project that then label your images. This will yield a labeled data set that can be leveraged to train a model to auto-clasify incoming images to Catalog. You could build your own model, or leverage Model inside Labelbox.
I’ve achieved something similar but for a different raw file type and a more diverse classification system.
Would be interesting to know a little more about your use-case here, namely, what do you hope to achieve by validating a screenshot image includes a Google Logo?