GUI for annotation of molecule images.
While there are several applications and web services available for drawing molecules, many struggle to recognize chemical structures from images, especially when those structures are complex. Our application, MOLInA, is designed to help bridge this gap. By enabling users to annotate molecule images, MOLInA provides essential data that can be used to enhance the accuracy of machine learning models through transfer learning. This tool is particularly useful for improving the recognition of moderately complex to complex molecules, which are currently not well handled by existing models.
In MOLInA, you can:
- 🖼️ Load images for annotation
- 🔬 Predict chemical structures using MolScribe (and soon, DECIMER)
- ➕ Add, 🔄 replace, or ❌ delete atoms and bonds
- 🧽 Clear annotations entirely
- ⌨️ Customize hotkeys
- 📄 Export to molfile format (currently in progress).
Coming soon:
- 🔋 Ability to add charge to molecules
- 🔄 DECIMER integration for improved predictions
- 📂 Batch processing for predicting structures in all images within a directory
- 💾 Options to save annotations in several styles.
Here's a look at the main interface of MOLInA:
When you open image
When you add atoms and bonds
Choose model
Show (you can choose) recently opened image
Execute > python -m molina
Later you can download .exe to run application
Clone repo to your computer and install via pip locally:
> git clone [email protected]:XenaN/MOLInA.git
> cd MOLInA
> pip install -e .After that, download and place MolScribe weights file to models directory.
MOLInA is licensed under the GNU General Public License (GPL). This means that anyone is free to use, modify, and redistribute the software under the terms of the GPL.