Page 175 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

164 Chapter 6 Plant leaf disease classification based on feature selection

Figure 6.8 VGG16 architecture.

A downside to VGG16 is that there are many more parameters
to be trained, leading to a significant training time.

3.3.3 ResNet (2015)
ILSVRC. The salient feature of ResNet was the residual block
technique, which allowed the neural network to be deeper while
keeping the number of parameters moderate.
With the advent of DL in various areas of research, important
applications were found. The major drawback of DL networks is
the size of data set required to train them along with the hyper-
parameters associated with them. Other issues include losing
the learning capability due to extensive stacking of layers and
overfitting. The research on residual networks was premised
on tackling the problem of vanishing gradients and training
efficiency between layers. The residual blocks provide mappings
that preserve the transformations in the previous layers.