Reading handwritten numbers using Deep Networks with MNIST Data in R Part3

Question

I try to write a program based on Deep Networks to read handwritten numbers. I found a code in Youtube (https://www.youtube.com/watch?v=5bso_5X7Zu4) which works there but it does not work for me. The problem is that I get error when I try to predict my handwritten number (namely number 5) that I have made in Windows Paint.

My number in Paint which its file name is number5.jpg is:

My complete code is:

library("pkgdown")
library(keras)
# devtools::install_github("rstudio/reticulate")
mnist <- dataset_mnist()
# str(mnist)
trainx <- mnist$train$x
trainy <- mnist$train$y
testx  <- mnist$test$x
testy  <- mnist$test$y

table(mnist$train$y, mnist$train$y)
table(mnist$test$y,  mnist$test$y)

# plot images
windows()
par(mfrow = c(3,3))
for (i in 1:9) plot(as.raster(trainx[i,,], max=255))
trainx[2,,]
windows()
hist(trainx[1,,])

# Reshape & rescale
trainx <- array_reshape(trainx, c(nrow(trainx), 784))
testx  <- array_reshape(testx,  c(nrow(testx),  784))
trainx <- trainx / 255
testx <-  testx / 255
#windows()
#hist(trainx[1,])
# One hot encoding
trainy <- to_categorical(trainy, 10)
testy  <- to_categorical(testy, 10)

# trainx <- as.matrix(trainx)
# trainy <- as.matrix(trainy)

# Model
model <- keras_model_sequential()
model %>% 
          layer_dense(units = 128, activation = 'relu', input_shape = c(784)) %>%
          layer_dropout(rate = 0.3) %>%
          layer_dense(units = 64,  activation = 'relu') %>%
          layer_dropout(rate = 0.2) %>%
          layer_dense(units = 10, activation = 'softmax')
summary(model)

# Compile
model %>%
               compile(loss ='categorical_crossentropy',
                       optimizer = optimizer_rmsprop(),
                       metrics = 'accuracy')
# Fit model
history <- model %>% 
            fit(trainx,
                trainy,
                epochs = 30,
                batch_size = 32,
                validation_split = 0.2)
plot(history)
# Evaluation and Precition - Test data
model %>% evaluate(testx, testy)

pred <- model %>% predict(testx) %>% k_argmax() %>% as.integer() %>% .[1:7840000]  

prob <- model %>% predict(testx)
cbind(Predicted_class = pred , Actual = mnist$test$y)[1:150,]

# New data
#install.packages("BiocManager") 
#BiocManager::install("EBImage")
library(EBImage)
setwd("C:/Users/hofo/Arbeidsmapper/Documents/NLP/NLP_BOSTOTTE/JPG_filer")

mypic <- readImage("number5.jpg")
mypic <- resize(mypic, 28, 28) 
mypic <- array_reshape(mypic, c(28, 28, 3))  


new <- NULL
new <- rbind(new, mypic)
str(new)
newx <- new[1:1,1:784]
newy <- c(5)

pred <- model %>% predict(newx) %>% k_argmax() %>% as.integer() %>% .[1:9]

The error when I run the last row is:

Error in py_call_impl(callable, dots$args, dots$keywords) : ValueError: in user code:

C:\Users\hofo\AppData\Local\R-MINI~1\envs\R-RETI~1\lib\site-packages\keras\engine	raining.py:1586 predict_function  *
    return step_function(self, iterator)
C:\Users\hofo\AppData\Local\R-MINI~1\envs\R-RETI~1\lib\site-packages\keras\engine	raining.py:1576 step_function  **
    outputs = model.distribute_strategy.run(run_step, args=(data,))
C:\Users\hofo\AppData\Local\R-MINI~1\envs\R-RETI~1\lib\site-packages	ensorflow\python\distribute\distribute_lib.py:1286 run
    return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
C:\Users\hofo\AppData\Local\R-MINI~1\envs\R-RETI~1\lib\site-packages	ensorflow\python\distribute\distribute_lib.py:2849 call_for_each_replica
    return self._call_for_each_replica(fn, args, kwargs)
C:\Users\hofo\AppData\Local\R-MINI~1\envs\R-RETI~1\lib\site-packages	ensorflow\python\distribute\distribute_lib.py:3632 _call_for_each_replica
    return fn(*args, **kwargs)

Can you also please help me with this?

Reading handwritten numbers using Deep Networks with MNIST Data in R Part3

0 Answers