clip

Compute image-text similarity scores with pretrained CLIP models.

Updated Apr 30, 2026

Search Tags:#retrieval #zero-shot #embedding #clip #vision-language #image-classification #image-text

Ready to run

npx skills add https://github.com/photonics-dhl/Hermes --skill clip-photonics-dhl

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: clip
Source: https://github.com/photonics-dhl/Hermes/tree/main/hermes-home/skills/mlops/models/clip
Command: npx skills add https://github.com/photonics-dhl/Hermes --skill clip-photonics-dhl

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

CLIP enables cross-modal understanding between images and text, enabling zero-shot classification and image-text retrieval without the need for fine-tuning.

Core Features & Use Cases

Zero-shot image classification and image-text similarity
Cross-modal retrieval, search, and content moderation
Visual question answering and image-to-text alignment

Quick Start

Provide an image and a set of text prompts and CLIP will return similarity scores for zero-shot classification.

Frequently Asked Questions about clip

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I perform zero-shot image classification without fine-tuning?▼

Zero-shot image classification without fine-tuning is achieved by computing image and text embeddings using pretrained CLIP models. You provide an image alongside text prompts, and the model returns similarity scores to classify the image.

What is CLIP and how does cross-modal image-text retrieval work?▼

CLIP is a vision-language model that bridges visual and textual understanding to perform cross-modal image-text retrieval. It works by loading pretrained models to compute and normalize image and text features for accurate similarity computations.

Can I use pretrained CLIP models for visual question answering and content moderation?▼

Yes, pretrained CLIP models support visual question answering and content moderation through image-to-text alignment. The Skill computes normalized embeddings to evaluate semantic similarity across diverse models and datasets.

How do I calculate image-to-text similarity scores for semantic search?▼

To calculate image-to-text similarity scores for semantic search, you load a pretrained CLIP model to generate embeddings. The Skill then normalizes these features and performs efficient similarity computations between the visual and textual inputs.

Does zero-shot image classification work with diverse datasets and model architectures?▼

Zero-shot image classification is designed to work across diverse models and datasets. The Skill applies cross-modal understanding to perform image-text retrieval and similarity computations without requiring task-specific training data.