Main approach uses a RCNN type model to directly predict individual atom and bond locations, and constructs molecules from those.

Note: Miniconda3 must be installed to install RDkit. The code runs on linux (tested on Ubuntu 16.04). Code tested on cuda10.1+pytorch1.6.

pip install requirements.txt
pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu101/torch1.6/index.html
conda install -c conda-forge rdkit=2020.03.1 

• main.py : Main entry to code, constructs trainer object to train model, and predicts smiles for images in test folder.
• trainer.py : Trainer class, preprocesses data, and trains model.
• labels_generation.py : Functions to preprocess data and generate annotations for RCNN.
• inference.py : Constructs mol object from predicted atom and bond bounding boxes .
• utils.py : Other general helper functions.

Set train_new_model to True in main.py and run main.py.

Set test_data_path in main.py to the directory containing test images and run main.py. Really important to set the parameter 'input_format' to "RGB" or "GBR" depending on the particular set of images!

No external data was used for any training or prediction, all was calculated directly from RDKit. Even the training images are generated in case they are not present in the folder "./data/images/train".

A pretrain model is available to obtain predictions similar to the competition submission. "./trained_models/final_submission.pth".

Due to random initialization difference between different operating systems, compute environment and GPU types, results can slightly differ between machines.