基于扩散模型多模态提示的电力人员行为图像生成

朱志航,闫云凤,齐冬莲

Image generation for power personnel behaviors based on diffusion model with multimodal prompts

Zhihang ZHU,Yunfeng YAN,Donglian QI

表 2 不同方法在多种指标下的定量结果

Tab.2 Quantitative results of different methods under multiple metrics

方法 FID KID CLIP-Score PCK/% OKS ControlNet[18] 274.72 6.62 67.33/30.31 47.5 0.872 HumanSD[20] 331.05 11.65 62.21/29.24 89.4 0.946 PoseNet 130.79 5.12 87.23/30.32 75.4 0.889 PoseNet+图像滤波器 130.43 4.90 89.81/30.43 90.4 0.978 PoseNet+图像滤波器+双阶段训练 128.25 4.56 91.02/31.44 94.2 0.979