Image Classification on ImageNet 1k (train) - Epoch Progression Metrics
69.2Accuracy @ 100 EpochsMatrix-SSL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Matrix-SSLBackbone=ResNet-50, Batch Size=512, Evaluation Protocol=Linear evaluation2026.03 | 69.2 | 69.9 | 71.1 | 71.9 | |
| IE-CLBackbone=ResNet-50, Batch Size=256, Evaluation Protocol=Linear evaluation2026.03 | 68.3 | 70.9 | 71.7 | 73.2 | |
| SimSiamBackbone=ResNet-50, Batch Size=256, Evaluation Protocol=Linear evaluation2026.03 | 68.1 | 70 | 70.8 | 71.7 | |
| MoCo-v2Backbone=ResNet-50, Batch Size=256, Evaluation Protocol=Linear evaluation2026.03 | 67.4 | 69.9 | 70.9 | 71.3 | |
| SimCLRBackbone=ResNet-50, Batch Size=4096, Evaluation Protocol=Linear evaluation2026.03 | 66.5 | 68.3 | 69.8 | 71.1 | |
| SwAVBackbone=ResNet-50, Batch Size=4096, Evaluation Protocol=Linear evaluation2026.03 | 66.5 | 69.1 | 70.7 | 71 | |
| NNCLRBackbone=ResNet-50, Batch Size=1024, Evaluation Protocol=Linear evaluation2026.03 | 65.4 | 66.1 | 66.8 | 68.7 | |
| All4OneBackbone=ResNet-50, Batch Size=1024, Evaluation Protocol=Linear evaluation2026.03 | 65.4 | 66 | 66.6 | 68.9 |