DECIMER V2 Models

The latest models for DECIMER V2.0

Update V2.4

- The latest models trained on TPU V4 and uses 512 x 512 as input images

- The fine-tuned model is available for hand-drawn images.

- The latest models can now understand chemical structure depictions with R-Groups and predict SMILES

- The model was trained on images generated using RanDepict 1.1.3

- DECIMER checkpoints also made available to use for further training

- Implementation of confidence score

For more details about implementation: https://github.com/Kohulan/DECIMER-Image_Transformer

Original Paper: 

• Rajan, K., Zielesny, A. & Steinbeck, C. DECIMER 1.0: deep learning for chemical image recognition using transformers. J Cheminform 13, 61 (2021). https://doi.org/10.1186/s13321-021-00538-8