Object recognition: apply *h5 cnn model to find an specific handwritten digit

[Leprechault]

I don't find any material about object detection using *h5 cnn adjusted model.
I try to create a simple example because I don't find in any book/web, I've like to apply any object detection technique to find handwritten digits in an image with several digits (wh_img). First, a create a CNN using the classic MNIST example:

Preparing the Data - using MNIST dataset is included with Kera

library(keras)
mnist <- dataset_mnist()
x_train <- mnist$train$x
y_train <- mnist$train$y
x_test <- mnist$test$x
y_test <- mnist$test$y

#reshape
x_train <- array_reshape(x_train, c(nrow(x_train), 784))
x_test <- array_reshape(x_test, c(nrow(x_test), 784))
#rescale
x_train <- x_train / 255
x_test <- x_test / 255

y_train <- to_categorical(y_train, 10)
y_test <- to_categorical(y_test, 10)

CNN model

model <- keras_model_sequential()
model %>%
layer_dense(units = 256, activation = 'relu', input_shape = c(784)) %>%
layer_dropout(rate = 0.4) %>%
layer_dense(units = 128, activation = 'relu') %>%
layer_dropout(rate = 0.3) %>%
layer_dense(units = 10, activation = 'softmax')

model %>% compile(
loss = 'categorical_crossentropy',
optimizer = optimizer_rmsprop(),
metrics = c('accuracy')
)

#Training and Evaluation
history <- model %>% fit(
x_train, y_train,
epochs = 30, batch_size = 128,
validation_split = 0.2
)
plot(history)

Save the model

model %>% save_model_hdf5("cnn_digits.h5")

Now, I've to apply any object detection technique in my y.png image with several handwritten digits to find only 4 numerals using my trained model :

Open an image with handwrite numbers collections

y = "http://mariakravtsova.us/img/numbers.png"
download.file(y,'y.png', mode = 'wb')
library("png")
wh_img <- readPNG("y.png")
plot.new()
rasterImage(wh_img,0,0,1,1)

library(image.darknet)

Try to use *h5 model to object detection only 4 handwritten digits

yolo_digits <- image_darknet_model(type = ‘detect’, labels ="4", model = "cnn_digits.h5")

x <- image_darknet_detect(file = "/Documents/numbers.png",
object = yolo_digits,
threshold = 0.4)

This is possible in R? Because in Python there are a lot of functions, choose the window size, etc. but in R not :(

[jwijffels]

You have build a model with keras and saved it as cnn_digits.h5. You can not pass such a model to work with darknet as that requires a .cfg file. Keras and darknet are 2 different frameworks.