Colin97/OpenShape_code

i'm confusing about the text feature with prompt_avg

alanzhangcs opened this issue · 2 comments

Thanks for your great work! I'm wondering could you release the generation process of text prompt_avg feature ?
For example prompt template ?

We follow https://github.com/salesforce/ULIP/blob/main/data/templates.json, to convert each category name to 64 sentences, and calculate the CLIP embedding for each sentence and then take their average as the embedding for the category name.

I got it. Thanks for your reply!