Open-Vocabulary Instance Segmentation on MARIS in-domain (val)
56.71Overall Class mAPMARIS
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| MARISBackbone=ConvNext-L2025.10 | 56.71 | 61.55 | 71.02 | 66.04 | 54.02 | 61.54 | 57.44 | 64.92 | 60.51 | |
| MAFT+Backbone=ConvNext-L2025.10 | 53.41 | 55.32 | 64.24 | 59.42 | 51.54 | 59.44 | 55.74 | 61.36 | 58.88 | |
| FCCLIPBackbone=ConvNext-L2025.10 | 52.17 | 54.29 | 63.33 | 58.37 | 50.99 | 58.66 | 54.57 | 60.33 | 55.92 | |
| EOVSegBackbone=ConvNext-L2025.10 | 49.53 | 51.72 | 63.16 | 55.57 | 48.32 | 57.26 | 51.53 | 59.36 | 53.04 | |
| SANBackbone=ViT-B, Group=High Performance Setting2025.10 | 47.78 | 50.17 | 59.63 | 55.08 | 46.05 | 53.47 | 49.76 | 55.86 | 51.92 | |
| ODISEBackbone=ViT-B, Group=High Performance Setting2025.10 | 46.95 | 49.32 | 58.75 | 54.26 | 45.18 | 52.64 | 48.93 | 55.02 | 51.07 | |
| OVSegBackbone=ViT-B, Group=High Performance Setting2025.10 | 46.41 | 48.96 | 57.92 | 53.64 | 44.63 | 51.89 | 48.25 | 54.23 | 50.36 | |
| MARIS (Our Method)Backbone=ConvNeXt-L2025.10 | 46.18 | — | — | — | — | — | — | 54.34 | 51.11 | |
| MARISBackbone=ConvNext-B2025.10 | 44.37 | 52.68 | 61.56 | 57.33 | 39.77 | 45.78 | 42.68 | 51.41 | 47.9 | |
| MAFT+Backbone=ConvNeXt-L2025.10 | 40.27 | — | — | — | — | — | — | 47.89 | 45.72 | |
| MAFT+Backbone=ConvNext-B2025.10 | 40.08 | 48.15 | 58.26 | 54.57 | 35.72 | 40.67 | 38.88 | 47.16 | 43.33 | |
| FCCLIPBackbone=ConvNeXt-L2025.10 | 39.46 | — | — | — | — | — | — | 46.39 | 43.62 | |
| FCCLIPBackbone=ConvNext-B2025.10 | 39.26 | 47.78 | 57.22 | 52.44 | 34.53 | 39.84 | 37.15 | 46.03 | 42.6 | |
| SANBackbone=ViT-B2025.10 | 36.05 | 43.26 | 52.18 | 48.05 | 31.57 | 37.09 | 34.02 | 43.06 | 39.26 | |
| EOVSegBackbone=ConvNeXt-L2025.10 | 35.9 | — | — | — | — | — | — | 45.33 | 40.11 | |
| ODISEBackbone=ViT-B2025.10 | 34.71 | 41.89 | 50.74 | 46.83 | 30.26 | 35.68 | 32.54 | 41.56 | 38.12 | |
| SANBackbone=ViT-B2025.10 | 34.05 | — | — | — | — | — | — | 42.2 | 38.26 | |
| ODISEBackbone=ViT-B2025.10 | 32.82 | — | — | — | — | — | — | 41.95 | 37.01 | |
| MARIS (Our Method)Backbone=ConvNeXt-B2025.10 | 32.62 | — | — | — | — | — | — | 39.6 | 36.65 | |
| EOVSegBackbone=ConvNext-B2025.10 | 31.22 | 37.98 | 48.95 | 41.55 | 27.48 | 33.89 | 29.56 | 39.26 | 33.83 | |
| OVSegBackbone=ViT-B2025.10 | 30.95 | 37.52 | 48.51 | 43.26 | 27.21 | 33.65 | 30.38 | 39.02 | 35.47 | |
| OVSegBackbone=ViT-B2025.10 | 30.65 | — | — | — | — | — | — | 40.78 | 37.9 | |
| MAFT+Backbone=ConvNeXt-B2025.10 | 30.05 | — | — | — | — | — | — | 36.57 | 34.11 | |
| FCCLIPBackbone=ConvNeXt-B2025.10 | 29.79 | — | — | — | — | — | — | 36.12 | 33.5 | |
| SANBackbone=ViT-B2025.10 | 19.18 | — | — | — | — | — | — | 24.63 | 20.05 | |
| OVSegBackbone=ViT-B2025.10 | 18.95 | — | — | — | — | — | — | 24.3 | 19.82 | |
| EOVSegBackbone=ConvNeXt-B2025.10 | 18.9 | — | — | — | — | — | — | 25.91 | 21.19 | |
| ODISEBackbone=ViT-B2025.10 | 18.51 | — | — | — | — | — | — | 23.86 | 19.4 |