infos.txt

Tip: Filling shapes can help!
Tip: Drawing context elements can help!
Tip: Big drawings are more likely to work!
Tip: Adding a face to objects doing actions helps!
Tip: Add a musical note to help CLIP guess instruments!
Tip: Add a fork to help CLIP guess food!
Did you know? CLIP has multimodal neurons!
Did you know? The paper introducing CLIP has been published in 2021.
Did you know? CLIP stands for 'Contrastive Language-Image Pretraining'.
Did you know? DALL·E 2 can generate images from text using CLIP's embeddings!
Did you know? Typographic attacks (adding text to images) can change CLIP's predictions!
Did you know? CLIP has been trained on 400 millions (image,text) pairs!