An extension of DECIMER V2 including the DECIMER Image Classifier — served as an API via Docker, local server, or standalone script.
Supports Mac Silicon (M1-4) GPU acceleration, Linux/Windows with CUDA, and reaction scheme image classification.
- Convert images of chemical structures to SMILES format
- Supports hand-drawn and digital images
- Classification also works with reaction schemes
- Fast recognition via server-based one-time model loading
- Easy Python API integration
docker compose up -d --buildThen install the Python client:
pip install ./packages/decimerapi/uv sync
uv run python decimer_server.pyThe server runs on localhost:8099. See example_usage for complete examples.
- Installation Guide - Docker, local, and standalone setup
- API Reference - HTTP and Python API usage
- Differences - Changes from original DECIMER packages
- FAQ - Common questions and important caveats
MIT License — see LICENSE file.