clip

Classify images with OpenAI's CLIP model for zero-shot recognition.

539|39|Updated May 1, 2026
One-click install
npx skills add https://github.com/Tommy-yw/RunbookHermes --skill clip-tommy-yw
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: clip
Source: https://github.com/Tommy-yw/RunbookHermes/tree/main/optional-skills/mlops/clip
Command: npx skills add https://github.com/Tommy-yw/RunbookHermes --skill clip-tommy-yw

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires transformers, torch, pillow, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill provides zero-shot image classification, image-text matching, and cross-modal retrieval capabilities, enabling users to perform tasks like image search, content moderation, and vision-language analysis without fine-tuning.

Core Features & Use Cases

  • Zero-shot Image Classification: Classify images without the need for training data.
  • Image-Text Matching: Match images with text descriptions.
  • Semantic Image Search: Search for images based on text queries.
  • Content Moderation: Detect inappropriate content in images.
  • Use Case: Use this Skill to automatically categorize and moderate user-generated content on a social media platform.

Quick Start

Use the clip skill to classify the image 'photo.jpg' into one of the following categories: 'a dog', 'a cat', 'a bird', 'a car', 'a person'.

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 labeled training data?

Zero-shot image classification without labeled training data uses OpenAI's CLIP model to match images against text descriptions. This Skill applies the CLIP model directly to categorize images based on semantic text prompts.

Can I use transformers and torch for image-text matching?

Yes, image-text matching is supported using the transformers and torch libraries. This Skill leverages these dependencies to compute similarity scores between image features and text descriptions via the CLIP model.

What is the best way to moderate user-generated images on a social media platform?

The best way to moderate user-generated images is using zero-shot vision-language models like CLIP. It detects inappropriate content by matching images against moderation text categories without fine-tuning.

Does semantic image search require fine-tuning the CLIP model?

No, semantic image search does not require fine-tuning the CLIP model. This Skill performs cross-modal retrieval directly, matching text queries against image embeddings to find relevant images out of the box.

How do I classify an image into custom categories using CLIP?

To classify an image into custom categories using CLIP, provide the image file and a list of text descriptions. The Skill processes the inputs with transformers and pillow to output the best matching category.