ResNet has been introduced in Deep Residual Learning for Image Recognition, Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun, 2015, https://arxiv.org/abs/1512.03385 . This network is very deep and can be trained using a standard stochastic descent gradient by using a standard network component called the residual module, which is then used to compose more complex networks (the composition is called network in network).
![](http://images-20200215.ebookreading.net/16/3/3/9781788293594/9781788293594__tensorflow-1x-deep__9781788293594__assets__ce5c2306-3a96-4a2b-86ff-b25f9faafdc5.png)
In comparison to VGG, ResNet is deeper, but the size of the model is smaller because a global average pooling operation is used instead of full-dense layers.