Discussion on image recognition under different conditions of Resnet and Densenet
Guangrui Li · Applied and Computational Engineering · 2023
As two mainstream and excellent neural networks, ResNet and DenseNet have been the main research directions of many scholars and papers. This paper mainly focuses on these two neural networks through a theoretical analysis, summary discussions, and practical experiments. The author explains their previous advantages and disadvantages, and, as a potentially obsolete technology, whether ResNet will be better or more efficient than DenseNet in some cases. The final experimental results show that although ReseNet is a somewhat outdated neural network model, it still outperforms DenseNet in some aspects. In addition, this paper provides some further optimization schemes for reference through the analysis of the two neural network models. At the same time, the basic structure and algorithm principle of the two neural networks are further explained in the analysis to make it clearer for readers.