Table 8. Performance comparison on brain hemorrhage classification dataset.
Pretext tasks indicated with two asterisks (**) are implemented using different backbones.
| No. | Author | Pretext task | Category | Random init.: Acc (%) | SSL: Acc (%) |
|---|---|---|---|---|---|
| 1 | Zhu et al. (2020b) | TCPC ∗∗ | Contrastive | 81.08 | 88.17 |
| 2 | Zhu et al. (2020a) | Rubik Cube+ ∗∗ | Predictive | 79.73 | 87.84 |
| 3 | Zhuang et al. (2019) | Rubik Cube | Predictive | 72.60 | 83.80 |
| 4 | Zhu et al. (2020a) | Rubik Cube+ ∗∗ | Predictive | 72.30 | 78.68 |
| 5 | Zhu et al. (2020b) | TCPC ∗∗ | Contrastive | 72.30 | 78.38 |