Image Captioning on COCO (CIDEr Drop Percentage)
2.18CIDEr Drop (%)Robust-LLaVA⁴_G
Evaluation Results
| Method | Links | |
|---|---|---|
| Robust-LLaVA⁴_GCorruption Type=JPEG2025.02 | 2.18 | |
| CLIPCorruption Type=Brightness2025.02 | 3.89 | |
| Sim-CLIP⁴Corruption Type=JPEG2025.02 | 7.05 | |
| Robust-LLaVA⁴_GCorruption Type=Brightness2025.02 | 7.17 | |
| CLIPCorruption Type=Pixelate2025.02 | 7.58 | |
| FARE⁴Corruption Type=JPEG2025.02 | 7.73 | |
| Robust-LLaVA⁴_GCorruption Type=Pixelate2025.02 | 8.65 | |
| CLIPCorruption Type=JPEG2025.02 | 10.66 | |
| Sim-CLIP⁴Corruption Type=Pixelate2025.02 | 12.98 | |
| CLIPCorruption Type=Snow2025.02 | 13.59 | |
| FARE⁴Corruption Type=Pixelate2025.02 | 14.15 | |
| CLIPCorruption Type=Fog2025.02 | 14.47 | |
| Sim-CLIP⁴Corruption Type=Brightness2025.02 | 16.45 | |
| CLIPCorruption Type=Frost2025.02 | 17.88 | |
| FARE⁴Corruption Type=Brightness2025.02 | 19.87 | |
| CLIPCorruption Type=Motion Blur2025.02 | 21.56 | |
| Robust-LLaVA⁴_GCorruption Type=Snow2025.02 | 25.35 | |
| Robust-LLaVA⁴_GCorruption Type=Elastic Transform2025.02 | 25.41 | |
| Robust-LLaVA⁴_GCorruption Type=Impulse Noise2025.02 | 25.67 | |
| CLIPCorruption Type=Defocus Blur2025.02 | 26.58 | |
| Robust-LLaVA⁴_GCorruption Type=Gaussian Noise2025.02 | 27.89 | |
| CLIPCorruption Type=Shot Noise2025.02 | 27.9 | |
| Robust-LLaVA⁴_GCorruption Type=Shot Noise2025.02 | 28.09 | |
| CLIPCorruption Type=Impulse Noise2025.02 | 28.66 | |
| Robust-LLaVA⁴_GCorruption Type=Frost2025.02 | 30.88 | |
| Sim-CLIP⁴Corruption Type=Elastic Transform2025.02 | 31.15 | |
| CLIPCorruption Type=Contrast2025.02 | 31.2 | |
| FARE⁴Corruption Type=Elastic Transform2025.02 | 32.62 | |
| CLIPCorruption Type=Elastic Transform2025.02 | 33.63 | |
| CLIPCorruption Type=Gaussian Noise2025.02 | 33.88 | |
| CLIPCorruption Type=Zoom Blur2025.02 | 34.32 | |
| Robust-LLaVA⁴_GCorruption Type=Motion Blur2025.02 | 39.56 | |
| Robust-LLaVA⁴_GCorruption Type=Zoom Blur2025.02 | 45 | |
| Robust-LLaVA⁴_GCorruption Type=Defocus Blur2025.02 | 45.9 | |
| Sim-CLIP⁴Corruption Type=Impulse Noise2025.02 | 46.62 | |
| Sim-CLIP⁴Corruption Type=Snow2025.02 | 47.68 | |
| Robust-LLaVA⁴_GCorruption Type=Glass Blur2025.02 | 48.56 | |
| Sim-CLIP⁴Corruption Type=Gaussian Noise2025.02 | 49.05 | |
| FARE⁴Corruption Type=Snow2025.02 | 50.21 | |
| Sim-CLIP⁴Corruption Type=Shot Noise2025.02 | 50.39 | |
| Sim-CLIP⁴Corruption Type=Frost2025.02 | 52.49 | |
| FARE⁴Corruption Type=Impulse Noise2025.02 | 52.99 | |
| FARE⁴Corruption Type=Shot Noise2025.02 | 53.26 | |
| FARE⁴Corruption Type=Gaussian Noise2025.02 | 55.67 | |
| Sim-CLIP⁴Corruption Type=Zoom Blur2025.02 | 55.91 | |
| Robust-LLaVA⁴_GCorruption Type=Fog2025.02 | 56.65 | |
| FARE⁴Corruption Type=Frost2025.02 | 57.62 | |
| FARE⁴Corruption Type=Zoom Blur2025.02 | 58.62 | |
| Sim-CLIP⁴Corruption Type=Motion Blur2025.02 | 58.69 | |
| CLIPCorruption Type=Glass Blur2025.02 | 58.91 | |
| Sim-CLIP⁴Corruption Type=Defocus Blur2025.02 | 59.7 | |
| FARE⁴Corruption Type=Defocus Blur2025.02 | 61.42 | |
| FARE⁴Corruption Type=Motion Blur2025.02 | 61.42 | |
| Sim-CLIP⁴Corruption Type=Glass Blur2025.02 | 69.85 | |
| FARE⁴Corruption Type=Glass Blur2025.02 | 70.07 | |
| Sim-CLIP⁴Corruption Type=Fog2025.02 | 82.02 | |
| FARE⁴Corruption Type=Fog2025.02 | 84.34 | |
| FARE⁴Corruption Type=Contrast2025.02 | 93.99 | |
| Sim-CLIP⁴Corruption Type=Contrast2025.02 | 95.69 | |
| Robust-LLaVA⁴_GCorruption Type=Contrast2025.02 | 95.73 |