Review of the ResNet-50 convolutional neural network architecture

Abstract

ResNet is a deep convolutional neural network architecture developed by the Microsoft Research team specifically for image processing. The primary feature of this architecture and its main advantage is the use of residual learning, which avoids learning at each step [8, 9]. This approach enables you to train networks with a large number of layers while avoiding the phenomenon of gradient decay, which occurs when the learning rate of a neural network slows down or stops at a significant number of epochs