Table 6.
Performance of MolScribe and DECIMER Image-Transformer on six clean benchmark datasets
| MolScribe | DECIMER | |||
|---|---|---|---|---|
| Pi | T | Pi | T | |
| JPO | 54.22% | 0.94 | 50.00% | 0.85 |
| CLEF | 74.34% | 0.96 | 63.54% | 0.89 |
| UOB | 87.67% | 0.99 | 69.97% | 0.95 |
| Hand_Drawn | 12.70% | 0.63 | 6.66% | 0.47 |
| Img2Mol_Test | 53.60% | 0.94 | 15.63% | 0.55 |
| Ingido | 39.41% | 0.98 | 12.24% | 0.62 |
| Average | 53.66% | 0.91 | 36.34% | 0.72 |
The performance is described as the proportion of occurrences of identical predictions Pi and the average Tanimoto similarity T.