Residual block

One layer in standard neural networks directly transfers analysis results to the next layer. But the most important feature that distinguishes residual blocks from standard neural networks is that just as one layer will feed the next layer directly, two or three layers will feed the next layer directly. Here, residual blocks perform jumping between layers in the logic of operation. The layers to be skipped here are those that contribute little to the performance of the model. Thus, residual blocks provide both performance contribution and time savings to the model. In standard convolutional models, when the number of consecutive layers increases, the depth of the model increases and consequently decreases the performance of the optimization methods used in the model. Layers deemed unnecessary with residual blocks contribute directly to the training of the model without any negativity in the optimization methods (Yue, Fu & Liang, 2018). The general structure of the residual blocks used in this study is shown in Fig. 5.