Referring Segmentation on MTRefSeg-21K v1 (val)
68.24mIoUMTRefSeg-R1
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| MTRefSeg-R1Vision Encoder=Swin-B, Text Encoder=Phi-1.5, Fine-tuned=Yes2026.05 | 68.24 | — | 67.82 | 73.9 | 69.52 | 63.51 | 54.81 | 32.39 | |
| RMSINVision Encoder=Swin-B, Text Encoder=BERT, Fine-tuned=Yes2026.05 | 66.51 | 68.18 | — | 73.31 | 67.77 | 59.97 | 49.81 | 24.53 | |
| LGCEVision Encoder=Swin-B, Text Encoder=BERT, Fine-tuned=Yes2026.05 | 66.25 | 68.88 | — | 72.6 | 67.28 | 60.38 | 49.82 | 25.8 | |
| FIANetVision Encoder=Swin-B, Text Encoder=BERT, Fine-tuned=Yes2026.05 | 65.82 | 67.03 | — | 72.27 | 66.91 | 59.85 | 49.39 | 24.22 | |
| RSRefSegVision Encoder=Swin-B, Text Encoder=BERT, Fine-tuned=Yes2026.05 | 64.99 | 60.55 | — | 70.99 | 64.9 | 57.78 | 48.63 | 26.87 | |
| LAVTVision Encoder=Swin-B, Text Encoder=BERT, Fine-tuned=Yes2026.05 | 64.76 | 67.98 | — | 70.87 | 64.93 | 57.81 | 47.65 | 24.36 | |
| RefSegformerVision Encoder=Swin-B, Text Encoder=BERT, Fine-tuned=Yes2026.05 | 64.07 | 66.67 | — | 71.05 | 65.21 | 57.77 | 46.75 | 23.8 | |
| SegEarth-R1Vision Encoder=Swin-B, Text Encoder=Phi-1.5, Fine-tuned=Yes2026.05 | 61.46 | — | 62.55 | 66.98 | 62.11 | 55.5 | 44.64 | 24.92 | |
| GSVA-13BVision Encoder=CLIP-L, Text Encoder=Vicuna-13B, Fine-tuned=Yes2026.05 | 54.81 | — | 52.81 | 58.89 | 49.14 | 39.34 | 30.55 | 21.37 | |
| GLAMM-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Fine-tuned=Yes2026.05 | 54.68 | — | 50.4 | 58.36 | 48.17 | 38.81 | 30.21 | 19.6 | |
| GSVA-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Fine-tuned=Yes2026.05 | 54.56 | — | 50.43 | 58.26 | 49.33 | 38.71 | 30.02 | 21.04 | |
| UniChangeVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Fine-tuned=Yes2026.05 | 54.42 | — | 49.44 | 57.88 | 50.91 | 42.84 | 31.84 | 16.38 | |
| CRIS-101Vision Encoder=ResNet-101, Text Encoder=ResNet-101, Fine-tuned=Yes2026.05 | 53.93 | — | — | 58.64 | 52 | 44.01 | 27.93 | 14.39 | |
| CRIS-50Vision Encoder=ResNet-50, Text Encoder=ResNet-50, Fine-tuned=Yes2026.05 | 53.46 | — | — | 58.37 | 51 | 41.63 | 25.01 | 13.56 | |
| LISA-13BVision Encoder=CLIP-L, Text Encoder=Vicuna-13B, Fine-tuned=Yes2026.05 | 53 | — | 51.16 | 56.53 | 45.92 | 36.89 | 28.24 | 19.88 | |
| LISA-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Fine-tuned=Yes2026.05 | 51.42 | — | 49.78 | 54.51 | 44.28 | 35.01 | 26.46 | 17.48 | |
| GSVA-13BVision Encoder=CLIP-L, Text Encoder=Vicuna-13B, Fine-tuned=No2026.05 | 16.25 | — | 17.14 | 10.18 | 7.06 | 4.08 | 2.16 | 0.53 | |
| UniGeoSegVision Encoder=Swin-B, Text Encoder=Phi-1.5, Fine-tuned=No2026.05 | 14.2 | — | 7.32 | 12.4 | 10.96 | 9.56 | 8.21 | 6.62 | |
| GeoPixel-7BVision Encoder=CLIP-L, Text Encoder=InternLM-XComposer2.5, Fine-tuned=Yes2026.05 | 14.06 | — | 13.71 | 7.79 | 6.02 | 4.58 | 3.62 | 3.03 | |
| GSVA-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Fine-tuned=No2026.05 | 13.53 | — | 14.02 | 5.81 | 3.55 | 2.74 | 1.49 | 0.53 | |
| GLAMM-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Fine-tuned=No2026.05 | 12.67 | — | 11.1 | 10.13 | 7.64 | 5.96 | 3.99 | 2.69 | |
| GeoPixel-7BVision Encoder=CLIP-L, Text Encoder=InternLM-XComposer2.5, Fine-tuned=No2026.05 | 10.8 | — | 12.35 | 5.46 | 3.62 | 2.2 | 1.46 | 0.86 | |
| SegEarth-R1Vision Encoder=Swin-B, Text Encoder=Phi-1.5, Fine-tuned=No2026.05 | 7.26 | — | 4.42 | 5.08 | 4.36 | 3.63 | 3.1 | 2.67 | |
| LISA-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Fine-tuned=No2026.05 | 5.84 | — | 7.32 | 1.68 | 1.01 | 0.43 | 0.14 | 0.05 | |
| LISA-13BVision Encoder=CLIP-L, Text Encoder=Vicuna-13B, Fine-tuned=No2026.05 | 2.46 | — | 2.62 | 0.34 | 0.24 | 0.24 | 0.24 | 0.24 |