Semantic Segmentation on ADE20K (train)
51.8mIoUdata2vec 2.0
Evaluation Results
| Method | Links | |
|---|---|---|
| data2vec 2.0Arch=ViT-L/16, Pre-training Data=IN-1k, Pre-training Epochs=200, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 51.8 | |
| MAEArch=ViT-L/16, Pre-training Data=IN-1k, Pre-training Epochs=1600, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 50.7 | |
| CrossMAEArch=ViT-L/16, Pre-training Data=IN-1k, Pre-training Epochs=800, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 49.6 | |
| BootlegArch=ViT-L/16, Pre-training Data=IN-1k, Pre-training Epochs=600, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 48.3 | |
| BootlegArch=ViT-B/16, Pre-training Data=IN-1k, Pre-training Epochs=600, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 46.6 | |
| CrossMAEArch=ViT-B/16, Pre-training Data=IN-1k, Pre-training Epochs=800, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 45 | |
| BootlegArch=ViT-S/16, Pre-training Data=IN-1k, Pre-training Epochs=600, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 44.3 | |
| MAEArch=ViT-B/16, Pre-training Data=IN-1k, Pre-training Epochs=1600, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 44 | |
| CrossMAEArch=ViT-S/16, Pre-training Data=IN-1k, Pre-training Epochs=800, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 42.2 | |
| data2vec 2.0Arch=ViT-S/16, Pre-training Data=IN-1k, Pre-training Epochs=200, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 40.5 | |
| MAEArch=ViT-S/16, Pre-training Data=IN-1k, Pre-training Epochs=800, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 39.2 | |
| I-JEPAArch=ViT-S/16, Pre-training Data=IN-1k, Pre-training Epochs=600, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 36.1 | |
| I-JEPAArch=ViT-L/16, Pre-training Data=IN-1k, Pre-training Epochs=600, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 31.9 | |
| I-JEPAArch=ViT-B/16, Pre-training Data=IN-1k, Pre-training Epochs=600, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 29.8 | |
| data2vec 2.0Arch=ViT-B/16, Pre-training Data=IN-1k, Pre-training Epochs=200, Fine-tuning Protocol=Full-FT, Training Data Percentage=100%2026.03 | 0.1 |