How to create wandb.Table with image previews for a big dataset with most efficiency?

I have a dataset that has ~40k images. I want to upload them all using arifacts, i used run.add_dir(“path_to_dir_with_images”).
I also have a csv that contains the labels of these images as well as the name of the image. Is there a way to create a wandb.Table so that i can explore this csv but also to have a column called images that has a reference to the images from the artifact via the file name. So that when i do EDA i can look at the images that are in the artifact instead of adding a column “images” to the csv and populating it with wandb.Image objects and then uploading it as a wandb.Table?

Hey Milos,

Thank you for bringing that up! I will be looking into it internally and let you know as soon as possible.

Hey Milos,

Thank you very much for your patience!
Yes, there is a way you can create a wandb.Table that references images in your artifact via their file names.

Hello Milos,

Is everything alright with your query? Have you tried it?

@bill-morrisson . I am interested as well. Could you provide a reference about how to do so?

1 Like

@bill-morrisson . Same here, I’m very interested in how to do this, as, in my case, the full table including wandb.Image objects does not fit in memory.

1 Like

Still no updates? Am also interested in this

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.