Method for improving performance based on CNN image recognition by using DCGAN

The invention discloses a method for improving performance based on CNN image recognition by using DCGAN. The method secondly combines an excellent data generation capability of the DCGAN with a framework based on the CNN image recognition, the DCGAN is an improved novel confrontation generation network on the basis of GAN, the method applies CNN to an original structure, the GAN have deep convolution characteristics, and a better feature representation form is obtained in the data generation aspect. The method solves the problems that training sample data is difficult to collect, the sample similarity is too large and the like in the image recognition process, the limitation of the sample quantity and quality on the classification model optimization problem is broken, the classification model is further strengthened, and the image recognition accuracy is improved. 本发明公开了种利用DCGAN提高基于CNN图像识别性能的方法，该方法将DCGAN出色的数据生成能力与基于CNN图像识别框架进行了二度结合，并且DCGAN是在GAN的基础上经过改进后的新型对抗生成网络，所述方法将CNN应用到了原始结构中，使得GAN具有了深度卷积的特性，并在数据生成方面拥有更好