If there is no text description for an image, what needs to be used as a substitute?

API
1

Hi Team,

Have a quick question. If there is no text description for an image, if I want to create text embeddings using OpenAI CLIP, should I be using nulls, empty strings, whitespaces or other methods as a substitute so the model will learn well on those embeddings?

Thank you!

1 Like
2

I have no idea, but pushing it just in case anyone else knows.