The network currently demonstrating the highest reported accuracy on the ImageNet ReaL dataset is Baseline (ViT-G/14).
Top Performers on ImageNet ReaL
ImageNet ReaL is a benchmark used to evaluate the performance of image classification models, providing a challenging and realistic assessment of their capabilities. The models are ranked based on their accuracy percentage, indicating how well they correctly classify images.
Here are the top-performing models on the ImageNet ReaL benchmark:
Rank | Model | Accuracy |
---|---|---|
1 | Baseline (ViT-G/14) | 91.78% |
2 | ViTAE-H (MAE, 512) | 91.20% |
3 | Model soups (ViT-G/14) | 91.20% |
4 | Meta Pseudo Labels (EfficientNet-B6-Wide) | 91.12% |
The Baseline (ViT-G/14) model leads the pack with an impressive accuracy of 91.78%. This signifies its strong capability in correctly identifying objects across a wide range of categories within the complex ImageNet ReaL dataset.