Cells image generation method based on VAE-SGAN
Kaikai Liu, Renjun Shuai, Li Ma, ZeXu
Abstract
Although it has entered the era of big data, the image data with important research value in the field of medical imaging is often scarce due to various reasons such as difficult acquisition. To increase the image data in these fields, people have developed computer simulations to generate images Methodological research. At present, there are two main types of image generation models with potential, namely, Variational Autoencoders (VAE) [1] and Generative Adversarial Networks (GAN) [2]. In the process of classifying white blood cells, there are often insufficient sample data and unbalanced data. How to effectively expand the sample size and solve the imbalance of training data has always been a hot issue in research. In order to solve this problem, this paper proposes an image generation method based on the fusion network of VAE and GANs, which avoids the shortcomings of the two generation methods when used alone, and makes the generated images have higher generation quality.