Text image generation method based on modulation fusion and generative adversarial network

The invention discloses a text image generation method based on modulation fusion and a comparative learning generative adversarial network, and the method comprises the following steps: building a modulation fusion module, designing a residual structure, comprising two text feature transformation layers of a main path, two convolution layers, and a convolution layer of a branch; a generator is established, and the generator is composed of a mapping network, eight modulation fusion modules, six up-sampling modules and a convolutional layer. A judger network structure is established, and the judger is composed of a feature extractor and three branches, wherein the three branches comprise a semantic reconstruction branch, an unconditional loss branch and a conditional loss branch. Establishing a comparative learning network to compare loss; and optimizing a loss function, wherein the loss function comprises generative adversarial loss and semantic reconstruction loss. According to the method, the image which is