CLIP

Connecting text and images trained by

Contrastive Learning with text and image at the same time. CLIP makes input text to

Embedding vector for image processing. This zero-shot capability made

Image Labeling requirement very low.