🧬 DECIMER.ai

Deep Learning for Chemical Image Recognition

Transform chemical structure images into machine-readable SMILES with state-of-the-art AI

🚀 Use DECIMER | 📖 Documentation | 💬 Discussions | 📄 Publications

🎯 Overview

DECIMER (Deep lEarning for Chemical IMagE Recognition) is an open-source, production-ready platform that revolutionizes chemical structure extraction from scientific literature. Powered by cutting-edge transformer-based deep learning, DECIMER automatically identifies, segments, and converts chemical structures into SMILES representations with remarkable accuracy.

🌟 Why DECIMER?

🧠 State-of-the-Art AI

_{Transformer architecture trained on millions of structures}

⚡ Production Ready

_{Battle-tested on thousands of scientific documents}

🔓 Open Source

_{MIT licensed for academic and commercial use}

🛠️ Self-Hosted

_{Complete control over your data and infrastructure}

✨ Core Capabilities

graph LR
    A[📄 PDF/Images] --> B[🔍 Segmentation]
    B --> C[🎯 Detection]
    C --> D[🧠 Recognition]
    D --> E[✅ SMILES]
    
    style A fill:#e3f2fd,stroke:#1976d2,stroke-width:2px
    style B fill:#fff3e0,stroke:#f57c00,stroke-width:2px
    style C fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px
    style D fill:#e8f5e9,stroke:#388e3c,stroke-width:2px
    style E fill:#d4edda,stroke:#155724,stroke-width:3px

🔥 Key Features

📑 Document Processing PDF Support: Extract structures from multi-page documents Image Formats: PNG, JPEG, WebP, HEIC support Batch Processing: Handle multiple files simultaneously High Resolution: Processes images at 300 DPI for optimal accuracy	🎨 Structure Recognition Printed Structures: Industry-standard depictions Hand-Drawn: Recognizes sketched molecules Complex Structures: Handles stereochemistry and large molecules Markush Detection: Identifies generic structures
🔍 Intelligent Segmentation Automatic Detection: Finds structures in complex layouts Pixel-Perfect Extraction: Maintains structure clarity Multi-Structure: Extracts all structures from a single image Classification: Distinguishes chemical from non-chemical images	🎯 Output & Validation SMILES Generation: Standard chemical notation InChIKey Creation: Unique molecular identifiers Validation: Automatic structure verification Interactive Editing: Built-in Ketcher editor for corrections

🚀 Quick Start

📋 Prerequisites

Requirement	Minimum	Recommended
💻 RAM	8 GB	16 GB+
💾 Storage	10 GB	20 GB+
🐳 Docker	Latest	Latest
🌐 Browser	Chrome 90+	Chrome/Edge Latest

⚡ Installation

🐧 Linux / macOS

# Clone the repository
git clone https://github.com/Steinbeck-Lab/DECIMER.ai
cd DECIMER.ai/
cp .env.example .env # Creates an environment file

# ⚠️ IMPORTANT: For systems with less than 32GB RAM
# Edit docker/app/supervisor.conf to reduce resource allocation
# See https://github.com/Steinbeck-Lab/DECIMER.ai/wiki for details

# Build and launch
docker compose build --no-cache
docker compose up -d

# Monitor startup (optional)
docker compose logs -f supervisor

🍎 For Apple Silicon (M1/M2/M3):

docker compose -f docker-compose.apple_silicon.yml build --no-cache
docker compose -f docker-compose.apple_silicon.yml up -d

🪟 Windows

Install Docker Desktop
Configure resources in Docker Desktop settings (4+ CPU cores, 8+ GB RAM)
Run as Administrator:

git clone https://github.com/Steinbeck-Lab/DECIMER.ai
cd DECIMER.ai\
cp .env.example .env

# Run the automated build script
build-windows.bat

Alternative manual approach:

docker-compose -f docker-compose.windows.yml build --no-cache
docker-compose -f docker-compose.windows.yml up -d

💡 Pro Tip: For better performance, consider using WSL2

🌐 Access Your Instance

Open your browser to http://localhost:80
Wait 5-10 minutes for model initialization ⏱️
Upload a PDF or image containing chemical structures
Download your results as SMILES strings and mol files! 🎉

📊 First-Time Setup: The initial startup loads several large neural network models. Subsequent starts will be much faster.

🏗️ Architecture

System Components

🔍 DECIMER Segmentation

_{Detects and extracts chemical structures from documents using Mask R-CNN}

📦 Repository • 📄 Paper

🧠 DECIMER Transformer

_{Converts structure images to SMILES using Vision Transformers}

📦 Repository • 📄 Paper

🎯 Image Classifier

_{Distinguishes chemical structures from other images with CNNs}

📦 Repository

🔧 Tech Stack

🎯 Use Cases

📚 Academic Research Literature data mining Chemical database curation Systematic reviews Patent analysis	🏭 Industry Applications High-throughput screening Competitive intelligence Legacy data digitization Regulatory documentation
🔬 Chemical Informatics Structure-activity relationships Chemical space exploration Property prediction pipelines Automated annotation	🎓 Education Creating digital resources Chemical structure databases Interactive learning materials Open educational resources

📊 Performance

Metric	Value	Details
🎯 Accuracy	>95%	On printed structures
⚡ Speed	~5s/structure	Including segmentation
📈 Scalability	1000s/day	With proper hardware
🔄 Formats	PDF, PNG, JPEG, WebP, HEIC	Multiple input types

📚 Documentation

Resource	Description
📖 Installation Guide	Detailed setup instructions for all platforms
🔧 Configuration	Customizing your DECIMER instance
🐛 Troubleshooting	Common issues and solutions
🚀 API Reference	Programmatic access guide
💡 Best Practices	Optimization tips and tricks

📖 Citation

If DECIMER.ai powers your research, please cite our work:

🏆 Primary Citation

@article{rajan2023decimer,
  title     = {DECIMER.ai: An open platform for automated optical chemical 
               structure identification, segmentation and recognition in 
               scientific publications},
  author    = {Rajan, Kohulan and Brinkhaus, Henning Otto and 
               Agea, Maria Inmaculada and Zielesny, Achim and 
               Steinbeck, Christoph},
  journal   = {Nature Communications},
  volume    = {14},
  number    = {1},
  pages     = {5045},
  year      = {2023},
  publisher = {Nature Publishing Group},
  doi       = {10.1038/s41467-023-40782-0}
}

📚 Additional Publications

Enhanced Architecture (2024)

@article{rajan2024advancements,
  title   = {Advancements in hand-drawn chemical structure recognition through 
             an enhanced DECIMER architecture},
  author  = {Rajan, Kohulan and Brinkhaus, Henning Otto and 
             Zielesny, Achim and Steinbeck, Christoph},
  journal = {Journal of Cheminformatics},
  volume  = {16},
  number  = {1},
  pages   = {78},
  year    = {2024},
  doi     = {10.1186/s13321-024-00872-7}
}

Segmentation Method (2021)

@article{rajan2021segmentation,
  title   = {DECIMER-Segmentation: Automated extraction of chemical structure 
             depictions from scientific literature},
  author  = {Rajan, Kohulan and Brinkhaus, Henning Otto and 
             Sorokina, Maria and Zielesny, Achim and Steinbeck, Christoph},
  journal = {Journal of Cheminformatics},
  volume  = {13},
  number  = {1},
  pages   = {20},
  year    = {2021},
  doi     = {10.1186/s13321-021-00496-1}
}

Transformer Architecture (2021)

@article{rajan2021transformer,
  title   = {DECIMER 1.0: deep learning for chemical image recognition 
             using transformers},
  author  = {Rajan, Kohulan and Zielesny, Achim and Steinbeck, Christoph},
  journal = {Journal of Cheminformatics},
  volume  = {13},
  number  = {1},
  pages   = {61},
  year    = {2021},
  doi     = {10.1186/s13321-021-00538-8}
}

Initial Research Paper (2020)

@article{rajan2020decimer,
  title   = {DECIMER: towards deep learning for chemical image recognition},
  author  = {Rajan, Kohulan and Zielesny, Achim and Steinbeck, Christoph},
  journal = {Journal of Cheminformatics},
  volume  = {12},
  number  = {1},
  pages   = {65},
  year    = {2020},
  doi     = {10.1186/s13321-020-00469-w}
}

🤝 Contributing

We welcome contributions from the community! Whether you're fixing bugs, adding features, or improving documentation, your help is appreciated.

🌟 Ways to Contribute

🐛 Report Bugs: Open an issue
💡 Suggest Features: Start a discussion
📝 Improve Docs: Submit pull requests for documentation
🔧 Fix Issues: Check out our good first issues
⭐ Star the Project: Show your support!

📋 Development Workflow

Fork the repository
Create a feature branch (git checkout -b feature/amazing-feature)
Make your changes with clear, descriptive commits
Test thoroughly
Push to your fork (git push origin feature/amazing-feature)
Open a Pull Request

See CONTRIBUTING.md for detailed guidelines.

💬 Community & Support

Get Help

💬 Discussions: For questions, ideas, and community interaction
🐛 Issues: For bug reports and feature requests
✉️ Email: For direct support and collaboration inquiries

📜 License

This project is licensed under the MIT License, making it free for both academic and commercial use.

MIT License

Copyright (c) 2025 Kohulan @ Steinbeck Lab

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

[Full license text in LICENSE file]

🏛️ About

🎓 Maintained by the Kohulan @ Steinbeck Group

Natural Products Cheminformatics Research Group
Institute for Inorganic and Analytical Chemistry
Friedrich Schiller University Jena, Germany

🔗 Our Ecosystem

Project	Description
🌴 COCONUT	Open Natural Products Database
🔍 DECIMER Segmentation	Structure Detection Library
🧠 DECIMER Transformer	Image-to-SMILES Model
🎯 DECIMER Classifier	Chemical Image Classification

📫 Connect With Us

⭐ Star History

🙏 Acknowledgments

Funded by Carl Zeiss Foundation and the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under the ChemBioSys (Project INF) - Project number: 239748522 - SFB 1127.

Made with ❤️ and ☕ for the global chemistry community

Democratizing access to chemical knowledge, one structure at a time

⬆ Back to Top

Name		Name	Last commit message	Last commit date
Latest commit History 229 Commits
app		app
bootstrap		bootstrap
config		config
database		database
docker		docker
logos		logos
public		public
resources		resources
routes		routes
storage		storage
tests		tests
.editorconfig		.editorconfig
.env		.env
.env.example		.env.example
.gitattributes		.gitattributes
.gitignore		.gitignore
.styleci.yml		.styleci.yml
LICENSE		LICENSE
README.md		README.md
artisan		artisan
build-windows.bat		build-windows.bat
caffeine_mod.png		caffeine_mod.png
composer.json		composer.json
composer.lock		composer.lock
docker-compose.apple_silicon.yml		docker-compose.apple_silicon.yml
docker-compose.windows.yml		docker-compose.windows.yml
docker-compose.yml		docker-compose.yml
package-lock.json		package-lock.json
package.json		package.json
phpunit.xml		phpunit.xml
server.php		server.php
tailwind.config.js		tailwind.config.js
vhost.conf		vhost.conf
webpack.mix.js		webpack.mix.js

License

Steinbeck-Lab/DECIMER.ai

Folders and files

Latest commit

History

Repository files navigation