Unsupervised Object Segmentation on COCO
36.7mBOiQASA
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| QASADec.=Trans., K-type=adaptive2026.01 | 36.7 | 45.5 | 35 | — | — | — | |
| CODADecoder type=Diffusion decoders2026.01 | 36.61 | 41.43 | — | 47.54 | 36.41 | 42.6 | |
| SPOT+FS+RCDec.=Trans., K-type=fixed2026.01 | 35.7 | 45.3 | — | — | — | — | |
| SlotAdaptDecoder type=Diffusion decoders2026.01 | 35.1 | 39.2 | — | 41.4 | 36.1 | 41.4 | |
| SPOTDec.=Trans., K-type=fixed2026.01 | 35 | 44.7 | 33 | — | — | — | |
| SPOT w/ ENSDecoder type=Autoregressive decoders2026.01 | 35 | 44.7 | — | 37.8 | 33 | — | |
| SPOT-MEvaluation Protocol=Max(Slot Attention, Decoder Masks), Multi-layer=true, MUFASA=true2026.02 | 34.8 | 45.5 | 32.5 | 35.6 | — | — | |
| SPOTEvaluation Protocol=Max(Slot Attention, Decoder Masks)2026.02 | 34.7 | 44.3 | 32.7 | 37.8 | — | — | |
| SPOT w/o ENSDecoder type=Autoregressive decoders2026.01 | 34.7 | 44.3 | — | 37.8 | 32.7 | — | |
| QASA-MLPDec.=MLP, K-type=adaptive2026.01 | 33.6 | 36.5 | 32.5 | — | — | — | |
| DINOSAUR-MEvaluation Protocol=Max(Slot Attention, Decoder Masks), Multi-layer=true, MUFASA=true2026.02 | 32.7 | 43 | 30.5 | 33.9 | — | — | |
| DINOSAUREvaluation Protocol=Max(Slot Attention, Decoder Masks)2026.02 | 31.6 | 39.7 | — | 34.1 | — | — | |
| DINOSAURDec.=Trans., K-type=fixed2026.01 | 31.6 | 39.7 | — | — | — | — | |
| DINOSAURDecoder type=Autoregressive decoders2026.01 | 31.6 | 39.7 | — | 34.1 | — | — | |
| DINOSAUR+FS+RCDec.=MLP, K-type=fixed2026.01 | 31.2 | 40.1 | — | — | — | — | |
| SlotDiffusionDecoder type=Diffusion decoders2026.01 | 31 | 35 | — | 37.2 | 31.2 | 36.5 | |
| Stable-LSDDecoder type=Diffusion decoders2026.01 | 30.4 | — | — | 35 | — | — | |
| MetaSlotDec.=MLP, K-type=adaptive2026.01 | 29.5 | — | 27.9 | — | — | — | |
| SLATEEvaluation Protocol=Max(Slot Attention, Decoder Masks)2026.02 | 29.1 | 33.6 | — | 32.5 | — | — | |
| SLATE2026.01 | 29.1 | 33.6 | — | — | — | — | |
| SLATEDecoder type=Autoregressive decoders2026.01 | 29.1 | 33.6 | — | 32.5 | — | — | |
| DINOSAUR-MLPDec.=MLP, K-type=fixed2026.01 | 27.7 | 30.9 | — | — | — | — | |
| DINOSAURDecoder type=MLP decoders2026.01 | 27.7 | 30.9 | — | 40.5 | — | — | |
| AdSlotDec.=MLP, K-type=adaptive2026.01 | 27.4 | — | — | — | — | — | |
| SAEvaluation Protocol=Max(Slot Attention, Decoder Masks)2026.02 | 17.2 | 19.2 | — | 21.4 | — | — | |
| SADecoder type=MLP decoders2026.01 | 17.2 | 19.2 | — | 21.4 | — | — |