ai-agent

perception_models: Unlock Cutting‑Edge Multimodal AI for Your Startup

Name: perception_models: Unlock Cutting‑Edge Multimodal AI for Your Startup
Author: facebookresearch

Leverage state‑of‑the‑art image/video CLIP and multimodal LLMs in ready‑to‑run Jupyter notebooks. Ideal for founders building AI‑driven products. 2.4k+ GitHub stars.

From the repo · “State-of-the-art Image & Video CLIP, Multimodal Large Language Models, and More!”

2,398 stars+12★ this week168 forksJupyter NotebookGuide quality 8/10 · reviewed Oct 6Updated 4/13/2026100% free · open source

What it does

Perception_models allows founders to leverage state-of-the-art image and video understanding models, including CLIP and multimodal large language models, for various applications.

When to use it

•When building an AI-powered image or video analysis tool
•When needing to integrate multimodal large language models into a startup's product
•When requiring advanced computer vision capabilities for a startup's application

Ready-to-paste prompt

python examples/image_classification.py --model clip --image path/to/image.jpg

Heads up: Ensure you have the necessary GPU resources and compatible CUDA version, as perception_models relies heavily on GPU acceleration for optimal performance.

Saves to your device

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

🛠️ Technical setup

Expect 20–40 minutes in a terminal — or let your AI agent drive it.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/perception-models && curl -fsSL https://workflowstacks.com/api/skills/perception-models/claude-skill -o ~/.claude/skills/perception-models/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

How perception_models: Unlock Cutting‑Edge Multimodal AI for Your Startup works

Codeflow

Free to inspect

Perception_models: Unlock Cutting‑Edge Multimodal AI for Your Startup is a large Jupyter Notebook project (~28k lines across 154 code files, plus 2 test files). It is a full software project: use it through its install path rather than reading it end to end. Last commit 6 months ago, Apache-2.0 license, has a test suite.

Size

Large codebase

~28k lines · 154 code files · ~4 h to skim

Setup

Developer setup

A real software project. Use it via its install path; don't expect to read it all.

Runs on

Python

No API keys detected

Jupyter Notebook 80%Python 18%Cuda 1%

Where to start reading

1
README.md
Start here — what it does and how to install it
2
requirements.txt
Dependencies and the commands it exposes
3
core/__init__.py
Inside core/ — the main logic begins here

What's in each folder

core/Core code — the actual logic47 files

apps/detection/Package77 files

apps/pe/Package47 files

apps/plm/Package26 files

READMEHas testsDocumentedNo CIApache-2.0 licenseLast update 6 mo ago

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

🛠️ Technical setup

Expect 20–40 minutes in a terminal — or let your AI agent drive it.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/perception-models && curl -fsSL https://workflowstacks.com/api/skills/perception-models/claude-skill -o ~/.claude/skills/perception-models/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

Quick Actions

Details

Creator

facebookresearch

Language

Jupyter Notebook