The classification of plankton is of critical importance for understanding marine ecological processes and global climate change. As most zooplankton are transparent, bright-field imaging and dark-field imaging are commonly used to observe stained and unstained zooplankton samples respectively. Different imaging methods and imaging devices present different imaging styles and image quality, which is not conducive to the uniform examination of data. We propose a method that enables the transformation of bright-field and dark-field images of medium-sized zooplankton based on a cycleconsistency against network Cycle-GAN model, which consists of two generative models G and two discriminative models (inspection models) D. The R, G,and B channel color information and one grey-scale channel information are input and the MSE loss and L1 loss functions are used to measure the loss of the generators and discriminators as well as the cyclic consistency to improve the image generation quality. After 200 rounds of training, the generated images closely resembled the real images, and the PSNR and SSIM were used as the evaluation metrics of the generated images. The results show that the PSNR of the generated images is below 0.3 and the SSIM is above 0.8, which means that the model can be used to maximize the preservation of the details of the original images.