gistfile1.txt

# following example from https://tensorflow.rstudio.com/tutorials/keras/classification

# do these steps first : https://tensorflow.rstudio.com/install/
library(keras)
library(tidyverse)
library(png)
library(magick)

# get images into R
imgs <- 
  list.files(path = "training-png/",
             full.names = TRUE)

# TODO: split into testing and training set

class_names <- 
  basename(imgs) %>% 
  str_remove_all(., "-|\\d|.png")

class_names_int <- 
  as.factor(class_names) %>% 
  as.integer()  %>% 
  array

# all the images need to have the same dimensions
# here we crop them from the centre point to 
# create a square image, each image has the same
# height and width in pixels

dim1_pixels <- 30
dim2_pixels <- 30

imgs_centre_crop <- 
map(imgs,
    ~magick::image_read(.x) %>% 
     magick::image_crop(
                        geometry=paste0(dim1_pixels, "x", dim2_pixels, "+0+0"),
                        gravity="center",
                        repage=TRUE) %>% 
      magick::image_convert( 
                            colorspace = "gray") 
    )

imgs_centre_crop_int <- 
  map(imgs_centre_crop, 
      ~image_data(.x, 'gray') %>% 
       as.integer() 
  )

# check does it look ok?
image_1 <- as.data.frame(imgs_centre_crop_int[[1]])
colnames(image_1) <- seq_len(ncol(image_1))
image_1$y <- seq_len(nrow(image_1))
image_1 <- gather(image_1, "x", "value", -y)
image_1$x <- as.integer(image_1$x)

ggplot(image_1, 
       aes(x = x, y = y, 
           fill = value)) +
  geom_tile() +
  scale_fill_gradient(low = "white", 
                      high = "black", 
                      na.value = NA) +
  scale_y_reverse() +
  theme_minimal() +
  theme(panel.grid = element_blank())   +
  theme(aspect.ratio = 1) +
  xlab("") +
  ylab("")

# convert to array and normalise values to fit in 0-1
imgs_centre_crop_int_array <- 
  simplify2array(imgs_centre_crop_int)

imgs_centre_crop_int_array <- 
  drop(imgs_centre_crop_int_array)

imgs_centre_crop_int_array_perm <- 
  aperm(imgs_centre_crop_int_array, 
        c(3, 1, 2))

imgs_centre_crop_int_array_perm_norm <- 
  imgs_centre_crop_int_array_perm / 255

# check does it look ok?
image_1 <- as.data.frame(imgs_centre_crop_int_array_perm_norm[1, , ])
colnames(image_1) <- seq_len(ncol(image_1))
image_1$y <- seq_len(nrow(image_1))
image_1 <- gather(image_1, "x", "value", -y)
image_1$x <- as.integer(image_1$x)

ggplot(image_1, aes(x = x, y = y, fill = value)) +
  geom_tile() +
  scale_fill_gradient(low = "white", high = "black", na.value = NA) +
  scale_y_reverse() +
  theme_minimal() +
  theme(panel.grid = element_blank())   +
  theme(aspect.ratio = 1) +
  xlab("") +
  ylab("")

par(mfcol=c(5,5))
par(mar=c(0, 0, 1.5, 0), xaxs='i', yaxs='i')
for (i in 1:25) {
  img <- imgs_centre_crop_int_array_perm_norm[i, , ]
  img <- t(apply(img, 2, rev))
  image(1:length(imgs), 1:length(imgs), img, 
        xaxt = 'n', yaxt = 'n',
        main = paste(class_names[i]))
}


model <- keras_model_sequential()
model %>%
  layer_flatten(input_shape = c(dim1_pixels, dim2_pixels)) %>%
  layer_dense(units = 128, activation = 'relu') %>%
  layer_dense(units = 10, activation = 'softmax')


model %>% compile(
  optimizer = 'adam',
  loss = 'sparse_categorical_crossentropy',
  metrics = c('accuracy')
)

model %>% fit(imgs_centre_crop_int_array_perm_norm, 
              class_names_int, 
              epochs = 5, 
              verbose = 2)


score <- 
  model %>% 
  evaluate(imgs_centre_crop_int_array_perm_norm, 
           class_names_int, 
           verbose = 0)
cat('Test loss:', score["loss"], "\n")

predictions <- model %>% predict(imgs_centre_crop_int_array_perm_norm)
par(mfcol=c(5,5))
par(mar=c(0, 0, 1.5, 0), xaxs='i', yaxs='i')
for (i in 1:25) {
  img <- imgs_centre_crop_int_array_perm_norm[i, , ]
  img <- t(apply(img, 2, rev))
  # subtract 1 as labels go from 0 to 9
  predicted_label <- which.max(predictions[i, ]) 
  true_label <- class_names[i]
  if (predicted_label == true_label) {
    color <- '#008800'
  } else {
    color <- '#bb0000'
  }
  image(1:dim1_pixels, 1:dim2_pixels, img, col = gray((0:255)/255), xaxt = 'n', yaxt = 'n',
        main = paste0(class_names[predicted_label], " (",
                      class_names[true_label], ")"),
        col.main = color)
}


# Grab an image from the test dataset
# take care to keep the batch dimension, as this is expected by the model
img <- imgs_centre_crop_int_array_perm_norm[1, , , drop = FALSE]
dim(img)
predictions <- model %>% predict(img)
predictions

# subtract 1 as labels are 0-based
prediction <- predictions[1, ] 
which.max(prediction)
# get name
class_names[which.max(prediction)]