Pixel-level Reasoning Segmentation on PRIST 1.0 (test)
14.72CIoUMIRAS
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MIRASEvaluation Protocol=Fine-tuning, Model Stage=Stage-22025.02 | 14.72 | 24.22 | 40.61 | 30.34 | |
| GPT-4oEvaluation Protocol=Zero-shot, Model Type=General MLLM2025.02 | 14.13 | 17.35 | 35.01 | 23.18 | |
| OMG-LLaVAEvaluation Protocol=Fine-tuning, Model Type=Segmentation-specific MLLM2025.02 | 13.84 | 21.54 | 49.31 | 29.98 | |
| MIRASEvaluation Protocol=Zero-shot, Model Stage=Stage-12025.02 | 13.12 | 15.64 | 45.11 | 23.22 | |
| LLaVA-v1.6-7BEvaluation Protocol=Zero-shot, Model Type=General MLLM2025.02 | 11.84 | 11.9 | 34.78 | 17.69 | |
| LLaVA-v1.5-7BEvaluation Protocol=Zero-shot, Model Type=General MLLM2025.02 | 11.25 | 11.35 | 25.9 | 15.68 | |
| LISAEvaluation Protocol=Fine-tuning, Model Type=Segmentation-specific MLLM2025.02 | 11.23 | 26.23 | 29.22 | 27.64 | |
| Qwen2-VL-7BEvaluation Protocol=Zero-shot, Model Type=General MLLM2025.02 | 10.64 | 10.3 | 15.88 | 12.5 | |
| LISAEvaluation Protocol=Zero-shot, Model Type=Segmentation-specific MLLM2025.02 | 10.45 | 15.33 | 43.07 | 15.09 | |
| PixelLMEvaluation Protocol=Fine-tuning, Model Type=Segmentation-specific MLLM2025.02 | 10.32 | 20.95 | 18.84 | 11.71 | |
| PixelLMEvaluation Protocol=Zero-shot, Model Type=Segmentation-specific MLLM2025.02 | 9.87 | 17.21 | 35.36 | 14.68 | |
| OMG-LLaVAEvaluation Protocol=Zero-shot, Model Type=Segmentation-specific MLLM2025.02 | 9.67 | 16.67 | 77.8 | 27.46 | |
| InternVL2-8BEvaluation Protocol=Zero-shot, Model Type=General MLLM2025.02 | 8.26 | 7.59 | 11.55 | 9.16 |