CryoFM: Foundation Model for Cryo-EM

Welcome to the CryoFM User Documentation!

CryoFM (Cryo-Electron Microscopy Foundation Model) is an advanced AI-driven toolkit for efficient, high-quality 3D cryo-EM density map processing. Designed to support cutting-edge research and practical applications in structural biology, CryoFM provides robust and flexible tools for tasks such as denoising, inpainting, super-resolution, and conditional generation of volumetric data.

---

Project Overview

CryoFM demonstrates the potential capabilities of foundation models for cryo-EM density analysis and offers diverse usage modes. Users can explore a range of applications – from fundamental restoration to advanced generative modeling – through command-line tools and Python APIs.

CryoFM currently includes two major versions:

• CryoFM2: An improved version validated on experimental data.
• CryoFM1: A model designed to demonstrate various capabilities in an idealized setting.

---

Key Features

• 3D Flow Matching Models trained directly in pixel space
• Ready-to-Use Sampling Pipelines for common and advanced cryo-EM inference tasks
• Flexible APIs for integration into custom pipelines
• Web-based Demo & Command-Line Tools for both beginners and power users

---

Get Started

• Installation Guide
• CryoFM2 Model Guide
• CryoFM1 Model Guide

Visit the User Guide for detailed documentation of CryoFM’s core modules, advanced features, and troubleshooting tips.

---

CryoFM is open-source and actively maintained.
We welcome feedback, suggestions, and contributions from the community!

For more information, please refer to the GitHub repository.