Image Classification on CIFAR-100 (test) (Accuracy and Ratio Metrics)
66.68AccuracySupervised (labels)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Supervised (labels)Backbone=MobileNetV2-0.35×, Number of Parameters=396K, Evaluation Protocol=Linear probe, Seeds=32026.05 | 66.68 | 100 | |
| CA-DSSLBackbone=MobileNetV2-0.35×, Number of Parameters=396K, Evaluation Protocol=Linear probe, Seeds=3, Loss Configuration=Recommended Lcls + Lms2026.05 | 62.7 | 94 | |
| SEEDBackbone=MobileNetV2-0.35×, Number of Parameters=396K, Evaluation Protocol=Linear probe, Seeds=32026.05 | 61.71 | 92.5 | |
| CA-DSSLBackbone=MobileNetV2-0.35×, Number of Parameters=396K, Evaluation Protocol=Linear probe, Seeds=3, Loss Configuration=full 3-loss2026.05 | 60.5 | — | |
| SimCLR-TinyBackbone=MobileNetV2-0.35×, Number of Parameters=396K, Evaluation Protocol=Linear probe, Seeds=32026.05 | 44.45 | 66.7 | |
| BYOL-TinyBackbone=MobileNetV2-0.35×, Number of Parameters=396K, Evaluation Protocol=Linear probe, Seeds=32026.05 | 5.29 | 7.9 | |
| Random InitBackbone=MobileNetV2-0.35×, Number of Parameters=396K, Evaluation Protocol=Linear probe, Seeds=32026.05 | 4 | 6 | |
| DINO-TinyBackbone=MobileNetV2-0.35×, Number of Parameters=396K, Evaluation Protocol=Linear probe, Seeds=32026.05 | 3.85 | 5.8 |