Spatial VQA on IaOR-VQA Quantitative
99.5Output AccuracyGPT-4o
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-4ospatial information integration=w ours2024.03 | 99.5 | 74.5 | 70.5 | 55 | |
| GPT-4Vspatial information integration=w ours2024.03 | 98.5 | 73.5 | 69.5 | 54.5 | |
| InstructBLIPspatial information integration=w ours2024.03 | 98.5 | 31.5 | 22 | 11.5 | |
| LLaVA-1.5spatial information integration=w ours2024.03 | 97 | 43 | 29 | 14.5 | |
| SpatialVLMBackbone=HF version2024.03 | 91 | 33.5 | 20.5 | 7.5 | |
| GPT-4ospatial information integration=w/o ours2024.03 | 31.5 | 14 | 8.5 | 3 | |
| InstructBLIPspatial information integration=w/o ours2024.03 | 28.5 | 8 | 2.5 | 0 | |
| LLaVA-1.5spatial information integration=w/o ours2024.03 | 23.5 | 16.5 | 6 | 2 | |
| GPT-4Vspatial information integration=w/o ours2024.03 | 0.8 | 0 | 0 | 0 |