An Improved Image Generation Algorithm Based on DCGAN
An improved image generation algorithm based on the traditional Deep Convolutional Genera-tive Adversarial Network(DCGAN)is proposed to address issues related to training stability and image quality.This enhancement leverages the Efficient Channel Attention(ECA)module,which efficiently rec-ognizes the importance of different channels in an image,thereby improving the stability of the training process and the quality of generated images.Additionally,Wasserstein distance is incorporated as part of the loss function to precisely measure the difference between the real and generated distributions.Further-more,a Focal Loss function is introduced to alleviate the instability associated with Jensen-Shannon(JS)divergence during optimization,thereby accelerating the model convergence.Extensive experimental re-sults demonstrate a significant improvement in image generation quality using the proposed algorithm.Spe-cifically,the generation accuracy reaches 97.6%,representing a 4.2%increase compared to the baseline DCGAN model.
万方数据
维普
