CLIP

CLIP, short for Contrastive Language-Image Pre-Training, is a neural network from OpenAI trained on a large corpus of image and text pairs. It uses a dual encoder design: a vision encoder and a text encoder map images and captions into a shared embedding space, where cosine similarity scores how well any caption describes any image. This enables zero-shot image classification, since arbitrary class names can be written as text prompts and matched against an image without task-specific training; OpenAI reported that CLIP matches the original ResNet-50's ImageNet accuracy in zero-shot evaluation without using any of the 1.28 million labeled training examples. The repository provides pretrained models such as ViT-B/32 with a simple PyTorch API exposing encode_image and encode_text functions, and the approach underpins image search, content filtering, and the text conditioning used by many diffusion models. Released under the MIT license, CLIP is a standard building block for researchers and engineers working on vision-language tasks.