def main():
num_classes = 10
x_train, y_train, x_test, y_test, input_shape = prepareMnistData(
0.1) #input_shape 28*28*1
model = createModel(input_shape, num_classes)
file = r'./weights/cnnTf2Kernel_2.h5'
if 1: #training
model.load_weights(file) #continue training
checkpointer = ModelCheckpoint(filepath=file,
verbose=0,
save_best_only=False)
history = model.fit(x=x_train,
y=y_train,
batch_size=800,
epochs=30,
callbacks=[checkpointer])
#printModelWeights(model)
loss = np.array(history.history['loss'])
acc = np.array(history.history['accuracy'])
#print('loss=',loss)
#print('acc=', acc)
#plotSubLossAndAcc(loss,acc)
else: #load from pretrained file
model.load_weights(file)
test_loss, test_acc = model.evaluate(x_test, y_test, verbose=2)
print('\nTest accuracy:', test_acc, 'loss=', test_loss)
def main():
num_classes = 10
x_train, y_train, x_test, y_test, input_shape = prepareMnistData(0.1)
print('x_train.shape = ',x_train.shape)
print('input_shape.shape = ',input_shape)
model = createModel(input_shape,num_classes)
model.fit(x_train, y_train, epochs=5)
def main():
num_classes = 10
x_train, y_train, x_test, y_test, input_shape = prepareMnistData(
0.2) #input_shape 28*28*1
model = createModel(input_shape, num_classes)
history = model.fit(x=x_train, y=y_train, epochs=50)
#printModelWeights(model)
test_loss, test_acc = model.evaluate(x_test, y_test, verbose=2)
print('\nTest accuracy:', test_acc, 'loss=', test_loss)
loss = np.array(history.history['loss'])
acc = np.array(history.history['accuracy'])
plotSubLossAndAcc(loss, acc)
def main():
x_train, y_train, x_test, y_test, input_shape = prepareMnistData(
0.1) #input_shape 28*28*1
model = createModel(input_shape, classes=10)
file = r'./weights/cnnTf2Kernel_2.h5' #r'./weights/cnnTf2Kernel.h5'
model.load_weights(file)
test_loss, test_acc = model.evaluate(x_test, y_test, verbose=2)
print('\nTest accuracy:', test_acc, 'loss=', test_loss)
#visualKernels(model)
#visualize_filter(model)
show32FilterImg(model)
def main():
x_train, y_train, x_test, y_test, input_shape = prepareMnistData(0.1) #input_shape 28*28*1
model = createModel(input_shape, classes=10)
file = r'./weights/cnnTf2Kernel_2.h5' #r'./weights/cnnTf2Kernel.h5'
model.load_weights(file)
test_loss, test_acc = model.evaluate(x_test, y_test, verbose=2)
print('\nTest accuracy:', test_acc,'loss=',test_loss)
testImg = x_test[6]
testLabel = y_test[6]
# print('testImg:', testImg.shape)
# print('testLabel:', testLabel)
# plt.imshow(testImg)
# plt.show()
visualModel(model, testImg)
def main():
x_train, y_train, x_test, y_test, input_shape = prepareMnistData(0.1) #input_shape 28*28*1
model = createModel(input_shape, classes=10)
file = r'./weights/cnnTf2Kernel_2.h5' #r'./weights/cnnTf2Kernel.h5'
model.load_weights(file)
test_loss, test_acc = model.evaluate(x_test, y_test, verbose=2)
print('\nTest accuracy:', test_acc,'loss=',test_loss)
testImg = x_test[6]
testLabel = y_test[6]
print('testLabel:', testLabel)
heatmap = getHeatMap(model, testImg)
# Display heatmap
plt.matshow(heatmap)
plt.show()
getCombineImg(testImg,heatmap, r'./res/7.png')