Referring Remote Sensing Image Segmentation on RRSIS-D (test)
71.91Mean IoU (mIoU)Earth-OneVision
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Earth-OneVisionScale=2B, Auxiliary=SAM22026.06 | 71.91 | — | — | — | — | — | 77.22 | — | — | |
| RemoteSAMScale=180M2026.06 | 71.75 | — | — | — | — | — | 80.04 | — | — | |
| RSRefSeg2Venue=TGRS’26, Paradigm=Fully-supervised specialist2026.07 | 69.17 | — | — | — | — | — | 79.45 | — | — | |
| GeoPixelScale=7B2026.06 | 67.99 | — | — | — | — | — | 81.77 | — | — | |
| Qwen3-VL-SAMLLM=Qwen3-VL-2B, Trained on RS data: LLM=LoRA, Trained on RS data: Mask Decoder=false, Trained on RS data: Extra=false2026.02 | 67.6 | — | — | — | — | — | — | — | — | |
| Sosa et al. (LoRA)Venue=arXiv’26, Paradigm=Trained vision–language model, Config=LoRA-tuned (Qwen3-VL-2B)2026.07 | 67.6 | — | — | — | — | — | — | — | — | |
| GeoPixelLLM=InternLM2-7B, Trained on RS data: LLM=LoRA, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 67.3 | — | — | — | — | — | — | — | — | |
| SegEarth-R1LLM=phi-1.5-1.3B, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 66.4 | — | — | — | — | — | — | — | — | |
| SegEarth-R1Scale=1.3B2026.06 | 66.4 | — | — | — | — | — | 67.56 | — | — | |
| SegEarth-R1Venue=arXiv’25, Paradigm=Trained vision–language model2026.07 | 66.4 | — | — | — | — | — | 78.01 | — | — | |
| BTDNetVenue=arXiv’25, Paradigm=Fully-supervised specialist2026.07 | 66.04 | — | — | — | — | — | 79.23 | — | — | |
| BTDNetLLM=BERT-base, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 66 | — | — | — | — | — | — | — | — | |
| SBANetLLM=BERT-base, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 65.5 | — | — | — | — | — | — | — | — | |
| ProVGVenue=arXiv’26, Paradigm=Fully-supervised specialist2026.07 | 65.44 | — | — | — | — | — | 77.62 | — | — | |
| RSRefSegLLM=SigLIP-So, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 64.7 | — | — | — | — | — | — | — | — | |
| CroBIMVenue=arXiv’25, Paradigm=Fully-supervised specialist2026.07 | 64.24 | — | — | — | — | — | 76.37 | — | — | |
| RMSINVisual Encoder=Swin-B, Text Encoder=BERT2023.12 | 64.2 | 74.26 | 67.25 | 55.93 | 42.55 | 24.53 | 77.79 | — | — | |
| RMSINLLM=BERT-base, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 64.2 | — | — | — | — | — | — | — | — | |
| FIANetVenue=TGRS’24, Paradigm=Fully-supervised specialist2026.07 | 64.01 | — | — | — | — | — | 76.81 | — | — | |
| FIANetLLM=BERT-base, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 64 | — | — | — | — | — | — | — | — | |
| ICIPNetVisual Encoder=Swin-B, Text Encoder=BERT2026.05 | 63.93 | 73.84 | 67.56 | 56.11 | 43.2 | 24.63 | 76.87 | — | — | |
| RMSINVenue=CVPR’24, Paradigm=Fully-supervised specialist2026.07 | 63.38 | — | — | — | — | — | 76.55 | — | — | |
| RemoteSAMVisual Encoder=Swin-B, Text Encoder=BERT2026.05 | 63.31 | 73.56 | 65.98 | 54.48 | 42.47 | 24.39 | 76.41 | — | — | |
| MAFNVisual Encoder=Swin-B, Text Encoder=BERT2026.05 | 63.21 | 73.35 | 67.2 | 56.28 | 42.61 | 23.62 | 76.14 | — | — | |
| FIANetVisual Encoder=Swin-B, Text Encoder=BERT2026.05 | 63.01 | 73.2 | 66.79 | 55.73 | 41.11 | 24.1 | 76.35 | — | — | |
| Text4Seg++Venue=TPAMI’26, Paradigm=Trained vision–language model2026.07 | 62.8 | — | — | — | — | — | 74.2 | — | — | |
| RMSINVisual Encoder=Swin-B, Text Encoder=BERT2026.05 | 62.37 | 71.7 | 65.1 | 53.92 | 41.25 | 22.46 | 75.91 | — | — | |
| DiffRISLLM=CLIP, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 62.2 | — | — | — | — | — | — | — | — | |
| LAVTVenue=CVPR’22, Paradigm=Fully-supervised specialist2026.07 | 61.12 | — | — | — | — | — | 76.48 | — | — | |
| LAVTVisual Encoder=Swin-B, Text Encoder=BERT2023.12 | 61.04 | 69.52 | 63.63 | 53.29 | 41.6 | 24.94 | 77.19 | — | — | |
| LAVTLLM=BERT-base, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 61 | — | — | — | — | — | — | — | — | |
| LGCEVenue=TGRS’24, Paradigm=Fully-supervised specialist2026.07 | 60.98 | — | — | — | — | — | 76.33 | — | — | |
| GeoGroundScale=7B, Auxiliary=SAM2026.06 | 60.5 | — | — | — | — | — | 61.1 | — | — | |
| GeoGroundVenue=arXiv’24, Paradigm=Trained vision–language model2026.07 | 60.5 | — | — | — | — | — | — | — | — | |
| RRSISLLM=BERT-base, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 59.4 | — | — | — | — | — | — | — | — | |
| LGCEVisual Encoder=Swin-B, Text Encoder=BERT2023.12 | 59.37 | 67.65 | 61.53 | 51.45 | 39.62 | 23.33 | 76.34 | — | — | |
| GeoSelectParadigm=Training-free2026.07 | 58.86 | — | — | — | — | — | 59.48 | — | — | |
| Earth-OneVisionScale=2B2026.06 | 52.82 | — | — | — | — | — | 64.12 | — | — | |
| RemoteReasonerScale=7B, Auxiliary=SAM22026.06 | 50.97 | — | — | — | — | — | 54.29 | — | — | |
| CMPC+Visual Encoder=R-101, Text Encoder=LSTM2023.12 | 50.24 | 57.65 | 47.51 | 36.97 | 24.33 | 7.78 | 68.64 | — | — | |
| LSCMVisual Encoder=R-101, Text Encoder=LSTM2023.12 | 49.92 | 56.02 | 46.25 | 37.7 | 25.28 | 8.27 | 69.05 | — | — | |
| BRINetVisual Encoder=R-101, Text Encoder=LSTM2023.12 | 49.65 | 56.9 | 48.77 | 39.12 | 27.03 | 8.73 | 69.88 | — | — | |
| CMPCVisual Encoder=R-101, Text Encoder=LSTM2023.12 | 49.24 | 55.83 | 47.4 | 36.94 | 25.45 | 9.19 | 69.22 | — | — | |
| CSMAVisual Encoder=R-101, Text Encoder=None2023.12 | 48.54 | 55.32 | 46.45 | 37.43 | 25.39 | 8.15 | 69.39 | — | — | |
| RRNVisual Encoder=R-101, Text Encoder=LSTM2023.12 | 45.64 | 51.07 | 42.11 | 32.77 | 21.57 | 6.37 | 66.43 | — | — | |
| PixelLMLLM=Vicuna-7B, Trained on RS data: LLM=LoRA, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 31.7 | — | — | — | — | — | — | — | — | |
| PixelLMScale=13B2026.06 | 31.65 | — | — | — | — | — | 33.89 | — | — | |
| RSVG-ZeroOVVenue=AAAI’26, Paradigm=Training-free2026.07 | 28.35 | — | — | — | — | — | 22.83 | — | — | |
| LISALLM=Vicuna-7B, Trained on RS data: LLM=LoRA, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 26.8 | — | — | — | — | — | — | — | — | |
| LISAScale=7B2026.06 | 26.78 | — | — | — | — | — | 27.84 | — | — | |
| NExT-ChatLLM=Vicuna-7B, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | 25 | — | — | — | — | — | — | — | — | |
| GPT-5-SAMLLM=GPT-5, Trained on RS data: LLM=false, Trained on RS data: Mask Decoder=false, Trained on RS data: Extra=false2026.02 | 24.9 | — | — | — | — | — | — | — | — | |
| Sosa et al. (zero-shot)Venue=arXiv’26, Paradigm=Training-free, Config=GPT-5 zero-shot2026.07 | 24.9 | — | — | — | — | — | — | — | — | |
| DGL-RSISVenue=JAG’26, Paradigm=Training-free2026.07 | 21.5 | — | — | — | — | — | — | — | — | |
| EKP-HRMVenue=GRSL’26, Paradigm=Training-free2026.07 | 19.21 | — | — | — | — | — | 20.68 | — | — | |
| GPT-Image-1LLM=GPT-5, Trained on RS data: LLM=false, Trained on RS data: Mask Decoder=false, Trained on RS data: Extra=false2026.02 | 17.2 | — | — | — | — | — | — | — | — | |
| SAM3 (sentence prompt)Venue=arXiv’26, Paradigm=Training-free2026.07 | 15.5 | — | — | — | — | — | 16.87 | — | — | |
| BRINetPublication=CVPR’202025.07 | — | 56.9 | 48.77 | 39.12 | 27.03 | 8.73 | — | 49.65 | 69.88 | |
| BTDNetPublication=Arxiv’252025.07 | — | 75.93 | 69.92 | 59.29 | 46.25 | 27.46 | — | 66.04 | 79.23 | |
| CADFormerPublication=JSTARS’252025.07 | — | 74.2 | 67.62 | 55.59 | 42.37 | 23.59 | — | 63.77 | 77.26 | |
| CMPCPublication=CVPR’202025.07 | — | 55.83 | 47.4 | 36.94 | 25.45 | 9.19 | — | 49.24 | 69.22 | |
| CMPC+Pub=TPAMI'21, Method Category=Segmentation Specialists2025.12 | — | — | — | — | — | — | — | 50.2 | — | |
| CMPC+Publication=TPAMI’212025.07 | — | 57.65 | 47.51 | 36.97 | 24.33 | 7.78 | — | 50.24 | 68.64 | |
| CMSAPublication=CVPR’192025.07 | — | 55.32 | 46.45 | 37.43 | 25.39 | 8.15 | — | 48.54 | 69.39 | |
| CroBIMPublication=Arxiv’242025.07 | — | 74.58 | 67.57 | 55.59 | 41.63 | 23.56 | — | 64.46 | 75.99 | |
| CrossVLTPublication=TMM’232025.07 | — | 70.38 | 63.83 | 52.86 | 42.11 | 25.02 | — | 61 | 76.32 | |
| EVF-SAMPublication=Arxiv’242025.07 | — | 72.16 | 66.5 | 56.59 | 43.92 | 25.48 | — | 62.75 | 76.77 | |
| FIANetPublication=TGRS’242025.07 | — | 74.46 | 66.96 | 56.31 | 42.83 | 24.13 | — | 64.01 | 76.91 | |
| GeoGroundPub=arXiv'24, Method Category=MLLM based segmentation2025.12 | — | — | — | — | — | — | — | 60.5 | — | |
| GeoPixelPub=ICML'25, Method Category=MLLM based segmentation2025.12 | — | — | — | — | — | — | — | 67.3 | — | |
| LAVTPub=CVPR'22, Method Category=Segmentation Specialists2025.12 | — | — | — | — | — | — | — | 61 | — | |
| LAVTPublication=CVPR’222025.07 | — | 69.52 | 63.63 | 53.29 | 41.6 | 24.94 | — | 61.04 | 77.19 | |
| LGCEPublication=TGRS’242025.07 | — | 67.65 | 61.53 | 51.45 | 39.62 | 23.33 | — | 59.37 | 76.34 | |
| LISAPub=CVPR'24, Method Category=MLLM based segmentation2025.12 | — | — | — | — | — | — | — | 26.8 | — | |
| LSCFPublication=TGRS’252025.07 | — | 74.3 | 67.69 | 56.32 | 43.08 | 25.67 | — | 64.25 | 77.42 | |
| LSCMPublication=ECCV’202025.07 | — | 56.02 | 46.25 | 37.7 | 25.28 | 8.27 | — | 49.92 | 69.05 | |
| PixelLMPub=CVPR'24, Method Category=MLLM based segmentation2025.12 | — | — | — | — | — | — | — | 31.7 | — | |
| RIS-DMMIPub=CVPR'23, Method Category=Segmentation Specialists2025.12 | — | — | — | — | — | — | — | 60.1 | — | |
| RIS-DMMIPublication=CVPR’232025.07 | — | 68.74 | 60.96 | 50.33 | 38.38 | 21.63 | — | 60.12 | 76.2 | |
| RMSINPub=CVPR'24, Method Category=Segmentation Specialists2025.12 | — | — | — | — | — | — | — | 64.2 | — | |
| RMSINPublication=CVPR’242025.07 | — | 74.26 | 67.25 | 55.93 | 42.55 | 24.53 | — | 64.2 | 77.79 | |
| RRNPublication=CVPR’182025.07 | — | 51.07 | 42.11 | 32.77 | 21.57 | 6.37 | — | 45.64 | 66.43 | |
| RS2-SAM 2Publication=Arxiv’252025.07 | — | 77.56 | 72.34 | 61.76 | 47.92 | 29.73 | — | 66.72 | 78.99 | |
| RSRefSeg 2Publication=-2025.07 | — | 80.23 | 75.78 | 65.41 | 50.65 | 31.05 | — | 69.17 | 79.45 | |
| RSRefSeg-lPublication=IGARSS’252025.07 | — | 74.49 | 68.33 | 58.73 | 48.5 | 30.8 | — | 64.67 | 77.24 | |
| SBANetPublication=Arxiv’252025.07 | — | 75.91 | — | 57.05 | — | 25.38 | — | 65.52 | 79.22 | |
| SegEarth-R1Pub=arXiv'25, Method Category=MLLM based segmentation2025.12 | — | — | — | — | — | — | — | 66.4 | — | |
| SegEarth-R1Publication=Arxiv’252025.07 | — | 76.96 | — | — | — | — | — | 66.4 | 78.01 | |
| SegEarth-R2Method Category=MLLM based segmentation2025.12 | — | — | — | — | — | — | — | 67.9 | — | |
| Text4Seg++Pub=arXiv'25, Method Category=MLLM based segmentation2025.12 | — | — | — | — | — | — | — | 62.8 | — |