hi everyone 
can anybody help me ??
I used this code before, it predicted the photos with a good approximation. But now it encounters an error. (Since I updated)
Can anyone tell me the correct coding?
I think the problem is in the last part of the coding
test <- lapply(mytest,readImage)
setwd("C:/Users/90553/Desktop/paste0")
train[[1]]
plot(test[[4]])
par(mfrow=c(7,3))
foreach(i=1:21) %do% {plot(train[[i]])}
par(mfrow=c(1,1))
foreach(i = 1:21) %do% {train[[i]] <-resize(train[[i]],150,150)}
foreach(i = 1:9) %do% {test[[i]] <-resize(train[[i]],150,150)}
str(test)
trainall <- combine(train)
testall <- combine(test)
display(tile(trainall,7))
str(trainall)
trainall <- aperm(trainall,c(4,1,2,3))
testall <- aperm(testall,c(4,1,2,3))
trainlabels <- rep(0:2, each=7)
testlabels <- rep(0:2 , each=3)
trainy <- to_categorical(trainlabels)
testy <- to_categorical(testlabels)
model <- keras_model_sequential()
model %>%
layer_conv_2d(filters = 256,
kernel_size=c(3,3),
activation = "relu",
input_shape = c(150,150,3))%>%
layer_max_pooling_2d(pool_size =c(3,3), strides = 2)%>%
layer_conv_2d(filters = 128,
kernel_size=c(5,5),
activation = "relu")%>%
layer_max_pooling_2d(pool_size = c(2,2), strides = 2)%>%
layer_conv_2d(filters = 32,
kernel_size=c(3,3),
activation = "relu")%>%
layer_max_pooling_2d(pool_size = c(2,2), strides = 2)%>%
layer_flatten()%>%
layer_dense(units = 256)%>% layer_activation_leaky_relu(alpha = 0.1)%>%
layer_dropout(rate = 0.4)%>%
layer_dense(units = 3, activation = "softmax")
summary(model)
model %>% compile(loss = "categorical_crossentropy",
optimizer = "rmsprop",
metrics = "accuracy")
model %>% compile(loss = "categorical_crossentropy",
optimizer = "rmsprop",
metrics = "accuracy")
history <- model %>%
fit(trainall,trainy, epoch=50,batch_size=32, validation_split=0.2)
save_model_hdf5(model, "model.cnn.h5")
model <- load_model_hdf5("model.cnn.h5")
train_evaluate <- evaluate(model, trainall, trainy)
test_evaluate <- evaluate(model, testall, testy)
model %>% predict(testall) %>% k_argmax()
pred <- testy
model %>% predict(testall) %>% k_argmax()
pred <- predict(testall)
pred[pred==0] <- "cat"
pred[pred==1] <- "dog"
pred[pred==2] <- "lion"
par(mfrow=c(3,3))
foreach(i=1:9) %do%{display(test[[i]], method = "raster");
text(x=20, y=20,label= pred[i],
adj = c(0,1),col = "black", cex = 4)}
par(mfrow=c(1,1))
