Extract data from json

[maurizius972]

Hello everyone. My question is foundamental for my thesis in computer science degree. I am using LabelBox to annotate some frame of video, i extract json of dataset and now i will have an info to ask. In the JSON i have the “embeddings”. I must insert in my work HOW LABELBOX EXTRACT THIS INFORMATION FROM DATASET. WHAT IS THE MODEL WHICH LABELBOX USE? Please help me

[ptancre]

Hi @maurizius972 ,

Nice that you are using Labelbox for your thesis !
As stated in our documentation we are using Gemini Pro vision to extract embedding from video data : Similarity search

Many thanks,

PT

[maurizius972]

you’re genius man, thanks

[maurizius972]

sorry i have a question. in the article that you suggest me, labelbox use [
CLIP-ViT-B-32](512 dimensions) to extract embedding from image. I am using labelbox to annotate a video but frame by frame. So, in my thesis, i must cite CLIP-ViT-B-32 or Google Gemini?

Can i have also your email? for future info, always it is possible

[ptancre]

You mentioned initially video so I suggested Google Gemini, now if you have extracted the video and doing annotation frame by frame (i.e. you have imported image frame in Labelbox then this would be CLIP-ViT-B-32).

The embeddings are computed depending on the media type.