Building the deep learning model

Once you have run the script from the preceding section, you should have 40,000 images for training in the cifar_10_images/data1/train directory and 8,000 images for validation in the cifar_10_images/data1/valid directory. We will train a model with this data. The code for this section is in the Chapter11/build_cifar10_model.R folder. The first section creates the model definition, which should be familiar to you:

library(keras)# train a model from a set of images# note: you need to run gen_cifar10_data.R first to create the images!model <- keras_model_sequential()model %>%  layer_conv_2d(name="conv1", input_shape=c(32, 32, 3),    filter=32, kernel_size=c(3,3), padding="same"  ) %>% layer_activation("relu") %>% ...

Get R Deep Learning Essentials now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.