Referring Image Segmentation on Geometry RIS benchmark
49IoUFlorence-2
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Florence-2Evaluation Protocol=fine-tuned2026.04 | 49 | 85 | |
| Qwen-VLEvaluation Protocol=fine-tuned2026.04 | 10 | 42 | |
| Qwen-VLEvaluation Protocol=zero-shot2026.04 | 3 | 5 | |
| LAVTPre-training dataset=RefCOCO, Evaluation Protocol=zero-shot2026.04 | 1 | 3 | |
| Florence-2Evaluation Protocol=zero-shot2026.04 | 1 | 3 |