Skip to main content

Evaluation Data


method →CNNSpot [1]FreDect [2]Fusing [3]Gram-Net [4]LGrad [5]LNP [6]DIRE-D [7]DIRE-G [7]UnivFD [8]PatchCraft
test setAccuracymAPAccuracymAPAccuracymAPAccuracymAPAccuracymAPAccuracymAPAccuracymAPAccuracymAPAccuracymAPAccuracymAP
ProGAN100.00 100.00 99.36 99.99 100.00 100.00 99.99 100.00 99.83 100.00 99.95 100.00 52.75 58.79 95.19 99.08 99.81 100.00 100.00 100.00
StyleGAN90.17 99.83 78.02 88.98 85.20 99.50 87.05 99.23 91.08 98.31 92.64 99.27 51.31 56.68 83.03 91.74 84.93 97.56 92.77 98.96
BigGAN71.18 85.99 81.97 93.62 77.40 90.70 67.33 81.79 85.63 92.93 88.43 94.54 49.70 46.91 70.13 75.25 95.08 99.27 95.80 99.42
CycleGAN87.62 94.94 78.77 84.78 87.00 95.50 86.07 95.33 86.94 95.01 79.07 89.52 49.58 50.03 74.19 80.56 98.33 99.80 70.17 85.26
StarGAN94.60 99.04 94.62 99.49 97.00 99.80 95.05 99.23 99.27 100.00 100.00 100.00 46.72 40.64 95.47 99.34 95.75 99.37 99.97 100.00
GauGAN81.42 90.82 80.57 82.84 77.00 88.30 69.35 84.99 78.46 95.43 79.17 84.54 51.23 47.34 67.79 72.15 99.47 99.98 71.58 81.33
StyleGAN286.91 99.48 66.19 82.54 83.30 99.60 87.28 99.11 85.32 97.89 93.82 99.70 51.72 58.03 75.31 88.29 74.96 97.90 89.55 97.74
whichfaceisreal91.65 99.85 50.75 55.85 66.80 93.30 86.80 95.21 55.70 57.99 50.00 42.75 53.30 59.02 58.05 60.13 86.90 96.73 85.80 95.26
ADM60.39 75.67 63.42 61.77 49.00 94.10 58.61 73.11 67.15 72.95 83.91 93.37 98.25 99.79 75.78 85.84 66.87 86.81 82.17 93.40
Glide58.07 72.28 54.13 52.92 57.20 77.50 54.50 66.76 66.11 80.42 83.50 92.76 92.42 99.54 71.75 78.35 62.46 83.81 83.79 94.04
Midjourney51.39 66.24 45.87 46.09 52.20 70.00 50.02 56.82 65.35 71.86 69.55 86.92 89.45 97.32 58.01 61.86 56.13 74.00 90.12 96.48
stable_diffusion_v_1_450.57 61.20 38.79 37.83 51.00 65.40 51.70 59.83 63.03 62.37 89.33 96.34 91.24 98.61 49.74 49.87 63.66 86.14 95.38 99.06
stable_diffusion_v_1_550.53 61.56 39.21 37.76 51.40 65.70 52.16 60.37 63.67 62.85 88.81 96.00 91.63 98.83 49.83 49.52 63.49 85.84 95.30 99.06
VQDM56.46 68.83 77.80 85.10 55.10 75.60 52.86 61.13 72.99 77.47 85.03 94.91 91.90 98.98 53.68 54.57 85.31 96.53 88.91 96.26
wukong51.03 57.34 40.30 39.58 51.70 64.60 50.76 55.62 59.55 62.48 86.39 95.33 90.90 98.37 54.46 55.38 70.93 91.07 91.07 97.54
DALLE250.45 53.51 34.70 38.20 52.80 68.12 49.25 49.82 65.45 82.55 92.45 98.26 92.45 99.71 66.48 74.48 50.75 63.04 96.60 99.56
sd-xl53.0372.6251.23 49.45 55.60 79.41 64.53 68.22 71.30 80.03 87.75 87.7591.28 99.10 55.35 53.97 50.7367.5998.4399.89
Average69.73 79.95 63.28 66.87 67.63 83.95 68.43 76.86 75.11 81.80 85.28 91.29 72.70 76.92 67.90 72.38 76.80 89.73 89.85 96.07

Citation

If you find this work useful for your research, please consider citing this bibtex.
@article{PatchCraft,
  title={PatchCraft: Exploring Texture Patch for Efficient AI-generated Image Detection},
  author={Zhong, Nan and Xu, Yiran and Qian, Zhenxing and Zhang, Xinpeng},
  journal={arXiv preprint arXiv:2311.12397},
  year={2023}
}