Multi-Human Image Generation on DiverseHumans TestPrompts (2-7 People)
92.4Count AccuracyDISCO
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DISCOBase Model=Flux, Source Type=Ours2025.10 | 92.4 | 98.6 | 98.3 | 33.4 | 85.6 | 81.7 | |
| GPT-Image-1Source Type=Proprietary2025.10 | 90.5 | 85.1 | 89.8 | 33.4 | 94.5 | 78.7 | |
| DISCOBase Model=Krea, Source Type=Ours2025.10 | 83.5 | 89.7 | 90.6 | 32.2 | 88.2 | 76.8 | |
| Qwen-ImageSource Type=Open-Source2025.10 | 79.8 | 49 | 45.9 | 32.6 | 93.3 | 60.1 | |
| Krea-DevSource Type=Open-Source2025.10 | 73.6 | 45.8 | 50.6 | 31.2 | 87.9 | 57.8 | |
| Gemini-NanobananaSource Type=Proprietary2025.10 | 72.3 | 57.2 | 42.7 | 31.9 | 95.7 | 60 | |
| Flux-DevSource Type=Open-Source2025.10 | 70.8 | 48.2 | 50.5 | 31.7 | 78.9 | 56 | |
| DreamOSource Type=Open-Source2025.10 | 70.5 | 31.8 | 35.2 | 32 | 82.7 | 50.4 | |
| OmniGen2Source Type=Open-Source2025.10 | 63.3 | 32.3 | 28.7 | 33.4 | 86.2 | 48.8 | |
| HiDreamSource Type=Open-Source2025.10 | 57.9 | 32.3 | 16.2 | 32.2 | 92.4 | 46.2 | |
| SD3.5Source Type=Open-Source2025.10 | 55.3 | 69.1 | 72.5 | 28.1 | 71.3 | 59.3 |