ChemGenius Tool
Chemical Structure Extractor
Upload any chemistry image or PDF and extract the SMILES string in seconds. Powered by DECIMER — a deep-learning optical chemical structure recognition model. No drawing required.
Predictions may be incorrect. Users must ALWAYS verify answers, calculations, and work.
Educational tool only. We are not responsible for misuse in academic settings.
Preparing document/image to SMILES workspace…
Use via API
Send a POST request with multipart/form-data to extract structures programmatically — useful for processing batches of documents or integrating into your research pipeline.
curl -X POST https://chemgenius.ai/api/calculators/organic/decimer-upload-to-smiles \ -H "Authorization: Bearer YOUR_API_KEY" \ -F "file=@structure.png" \ -F "max_structures=5"
Returns a list of candidate SMILES strings with optional confidence scores. Get API access →
How the extraction works
ChemGenius uses DECIMER (Deep lEarning for Chemical IMagE Recognition), a transformer-based model trained on millions of chemical structure images. When you upload a file, DECIMER locates and segments the structure regions, vectorizes the bond graph, and outputs one or more candidate SMILES strings ranked by confidence.
Each candidate is rendered as a 2D structure so you can visually confirm the result before using it downstream. For ambiguous or low-contrast scans, multiple candidates help you choose the most chemically sensible interpretation.
Upload
PDF, PNG, JPG, or WEBP — up to 20 MB
Recognize
DECIMER AI locates and reads the structure
Compare
Pick the best candidate from rendered 2D previews
Export
Copy SMILES into any downstream chemistry tool
Tips for best results
- Use high-resolution scans (300 dpi or above) for printed or textbook structures.
- Crop the image tightly around the target structure to reduce noise from surrounding text.
- Hand-drawn structures are supported but may produce lower confidence — clean up the image before uploading.
- For multi-structure documents, extract one structure at a time for cleaner output.
- Always verify extracted SMILES against a reference before using in research or lab protocols.
Frequently Asked Questions
What file types can I upload?
The extractor accepts PDF, JPG, JPEG, PNG, and WEBP files up to 20 MB. For best accuracy, upload a clean, high-resolution scan or a tightly cropped image focused on the target structure.
What is SMILES and why does it matter?
SMILES (Simplified Molecular Input Line Entry System) is the standard text notation for chemical structures. Once you have a SMILES string you can plug it directly into reaction predictors, databases like PubChem, and molecular drawing tools.
Why are multiple candidate structures returned?
Ambiguous scans — overlapping atoms, low contrast, hand-drawn bonds — can produce more than one plausible interpretation. Reviewing all candidates and comparing the rendered 2D structures helps you pick the most chemically accurate result.
How accurate is the extraction?
ChemGenius uses the DECIMER deep-learning model, which achieves strong accuracy on clean, typeset structures. Hand-drawn or heavily degraded images produce lower confidence. Always verify before using extracted SMILES in research or lab contexts.
Can I use extracted SMILES in other ChemGenius tools?
Yes. Copy the selected SMILES into Reaction Solver to predict products, Mechanism Predictor to trace an electron-pushing pathway, or the Organic Chemistry AI chat for in-depth analysis.
Is there an API for bulk structure extraction?
Yes. Send a POST request to /api/calculators/organic/decimer-upload-to-smiles with multipart/form-data and an Authorization header. See the API docs section below for a curl example.
Continue with your extracted structure
Got your SMILES? Take the next step directly in ChemGenius.