Computer Vision News - September 2019

We Tried for You 20 Here we can see that our training and testing error values are very close. This shows that we were able to generate some generalization by augmenting more data. Training the network on the full data set may produce even better results. Conclusion Data augmentation is the common practice when training a neural network. It enlarges the data set and enables to boost the generalization capabilities of the network. Keras data generator enables to augment data on the fly without taking care of all the technical details of the transformations. We implemented a network to show how it helps to generalize better when training a network. The network is quite simple and can run also on the CPU so you can try it at home. To solve the poor generalization property of the previews training, we use data augmentation. To this end, we generate data as we demonstrated at the beginning of this article. We will run a few more step at each epoch to enable the data generator to augment enough examples. The history of the training can be seen below: