Training Data Provenance Verification on ImageNet-100
78.7AccuracyTrainProVe
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TrainProVeBackbone=Swin-B2025.03 | 78.7 | 48.7 | 76.6 | |
| TrainProVeBackbone=ResNet182025.03 | 78.6 | 49.5 | 76.9 | |
| TrainProVeTt=33.6h, Tg=0.1h, Tv=1.4h, Total=35.1h2025.03 | 78 | 63 | 84 | |
| TrainProVeBackbone=ConvNeXt-B2025.03 | 77.1 | 47 | 75.6 | |
| Han et al.'s Work2025.03 | 67.1 | 42.5 | 64.5 | |
| TrainProVe-Sim2025.03 | 55.5 | 30.5 | 52.8 | |
| Random2025.03 | 50 | 28.6 | 50 | |
| Han et al.'s WorkTt=137.7h, Tg=8.0h, Tv=1.2h, Total=146.9h2025.03 | 48 | 28 | 48 | |
| TrainProVe-Ent2025.03 | 35.6 | 36.4 | 56.7 |