CLIP

Sigmoid Loss for Language Image Pre-Training (SigLIP)

SigLIP is an improved version of CLIP developed by Google DeepMind. It introduces sigmoid-based contrastive loss instead of the traditional softmax-based contrastive loss used in CLIP.

Resources

• https://arxiv.org/pdf/2303.15343
• https://huggingface.co/docs/transformers/en/model_doc/siglip