This video explains how CLIP from OpenAI transforms Image Classification into a Text-Image similarity matching task. This is done with Contrastive Training and Zero-Shot Pattern-Exploiting Training. Thanks for watching!
Paper Links:
Clip (Blog Post): openai.com/blog/clip/
VirTex: arxiv.org/pdf/2006.06666.pdf
ConVIRT: arxiv.org/pdf/2010.00747.pdf
Pattern-Exploiting Training: arxiv.org/pdf/2001.07676.pdf
Vision Transformer (Blog Post, Nice Animation): ai.googleblog.com/2020/12/tra...
Thanks for watching! Please Subscribe!
Негізгі бет Ғылым және технология CLIP: Connecting Text and Images
Пікірлер: 13