Visual Question Answering on Medical VQA DME dataset (test)
0.8349Overall AccuracyConsistency-preserving VQA
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Consistency-preserving VQAbaseline=Baseline (att.), lambda=0.5, gamma=12022.06 | 0.8349 | 0.8069 | 0.8496 | 0.8196 | 0.8316 | 0.942 | 0.9812 | |
| Baselineattention=yes2022.06 | 0.8146 | 0.8023 | 0.8313 | 0.8718 | 0.8058 | 0.8921 | 0.9692 | |
| SQUINTbaseline=Baseline (att.)2022.06 | 0.8058 | 0.7748 | 0.8282 | 0.8534 | 0.8002 | 0.8817 | 0.9462 | |
| Baselineattention=none2022.06 | 0.7754 | 0.7359 | 0.8137 | 0.8337 | 0.7666 | 0.817 | 0.9186 |