Panoptic Segmentation on ADE20K 150 (val)
44.46PQWOW-Seg
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| WOW-SegProtocol=Open-vocabulary, Evaluation Input=Ground truth box/mask2026.05 | 44.46 | — | — | |
| Osprey-7BProtocol=Open-vocabulary, Evaluation Input=Ground truth box/mask2026.05 | 41.89 | — | — | |
| Ferret-7BProtocol=Open-vocabulary, Evaluation Input=Ground truth box/mask2026.05 | 39.46 | — | — | |
| CLIP-ConvNeXt-LProtocol=Open-vocabulary, Evaluation Input=Ground truth box/mask2026.05 | 36.86 | — | — | |
| GPT4RoI-7BModel size=7B2023.07 | 36.32 | — | — | |
| GPT4RoI-7BProtocol=Open-vocabulary, Evaluation Input=Ground truth box/mask2026.05 | 36.32 | — | — | |
| RetCLIPBackbone=CLIP-ConvNeXt-large, Database=ADE20k, Fine-tuned on COCO=true2026.01 | 30.9 | 19.3 | 44 | |
| RetCLIPBackbone=CLIP-ConvNeXt-large, Database=Google Open Image, Fine-tuned on COCO=true2026.01 | 28.3 | 17.7 | 38.3 | |
| Shikra-7BConcurrent work=true, Model size=7B2023.07 | 27.52 | — | — | |
| Shikra-7BProtocol=Open-vocabulary, Evaluation Input=Ground truth box/mask2026.05 | 27.52 | — | — | |
| RetCLIPBackbone=CLIP-RN50x64, Database=ADE20k, Fine-tuned on COCO=true2026.01 | 27.1 | 15.1 | 36.9 | |
| CLIP-Surgery-ViT-LBackbone=ViT-L2023.07 | 26.55 | — | — | |
| CLIP-Surgery-ViT-LProtocol=Open-vocabulary, Evaluation Input=Ground truth box/mask2026.05 | 26.55 | — | — | |
| FC-CLIPBackbone=CLIP-ConvNeXt-large, Database=N/A, Fine-tuned on COCO=true2026.01 | 26.4 | 16.8 | 34 | |
| RetCLIPBackbone=CLIP-RN50x64, Database=Google Open Image, Fine-tuned on COCO=true2026.01 | 23.7 | 14.1 | 32.4 | |
| FC-CLIPBackbone=CLIP-RN50x64, Database=N/A, Fine-tuned on COCO=true2026.01 | 22.8 | 13.6 | 28.4 | |
| Kosmos-2Concurrent work=true2023.07 | 6.53 | — | — | |
| Kosmos-2Protocol=Open-vocabulary, Evaluation Input=Ground truth box/mask2026.05 | 6.53 | — | — |