Oracle Semantic Segmentation on PASCAL VOC 2012 (val)
82mIoUSpectral Clustering of Layer-Distributed Neural Representations + K-Means
Evaluation results
| Method | Links | |
|---|---|---|
| Spectral Clustering of Layer-Distributed Neural Representations + K-MeansBackbone=Stable Diff. 1.4, Affinity Source=All Attentions2023.12 | 82 | |
| Spectral Clustering of Layer-Distributed Neural Representations + K-MeansBackbone=CLIP VIT-B/16, Affinity Source=All Attentions2023.12 | 78 | |
| Spectral Clustering of Layer-Distributed Neural Representations + K-MeansBackbone=DINO VIT-S/16, Affinity Source=All Attentions2023.12 | 78 | |
| K-MeansBackbone=DINO VIT-S/16, Affinity Source=Final Features2023.12 | 74 | |
| Spectral Clustering of Layer-Distributed Neural Representations + K-MeansBackbone=MAE VIT-B/16, Affinity Source=All Attention2023.12 | 74 | |
| Normalized Cut (Ncut) + K-MeansBackbone=DINO VIT-S/16, Affinity Source=Final Features2023.12 | 73 | |
| MaskCutBackbone=DINO VIT-S/16, Affinity Source=Final Features2023.12 | 64 | |
| Normalized Cut (Ncut) + K-MeansBackbone=MAE VIT-B/16, Affinity Source=Final Features2023.12 | 62 | |
| Normalized Cut (Ncut) + K-MeansBackbone=DINO VIT-S/16, Affinity Source=Final Attentions2023.12 | 58 | |
| K-MeansBackbone=CLIP VIT-B/16, Affinity Source=Final Features2023.12 | 57 | |
| K-MeansBackbone=MAE VIT-B/16, Affinity Source=Final Features2023.12 | 48 | |
| Normalized Cut (Ncut) + K-MeansBackbone=CLIP VIT-B/16, Affinity Source=Final Features2023.12 | 45 |