Robustness test data generation method of pneumonia auxiliary diagnosis system

The invention provides a method for generating robustness test data of a pneumonia auxiliary diagnosis system, and the method comprises the steps: creating a CycleGAN network which is used for automatically generating test data of a pneumonia medical image with personalized interference through the definition of input parameters; the CycleGAN network comprises a generator and a discriminator, and the generator is used for outputting generated data Xfake = G (c, z) according to an input classification label c and random noise z; the discriminator is used for judging whether the generated data Xfake is a probability distribution P (SX) of real data and a probability distribution P (CX) of the generated data Xfake for a classification label; a CBAM attention module is introduced into the CyclGAN network and is used for replacing an original residual block in the network; a local discriminator for pneumonia is introduced into the CycleGAN network. According to the method, related test data can be generated instea