I defined my model starting with inputs = tf.keras.Input(shape=(512, 512, 3), batch_size=BATCH_SIZE). Then I use model.fit with data of shape (1, 720, 1280, 3). The model still trains normally and the loss decreases. Why is that?
Thanks.
I tried to train with same shape. But the result did not turns out to be good.
So here is the code:
inputs = tf.keras.Input(shape=(512, 512, 3), batch_size=BATCH_SIZE) # REVISE,
x = tf.keras.layers.Cropping2D(cropping=((256, 0), (0, 0)))(inputs)
x = tf.keras.layers.Resizing(66,200)(x)
# x = x / 255
x = tf.keras.layers.Conv2D(24, (5,5), (2,2), activation="elu")(x)
x = tf.keras.layers.Conv2D(36, (5,5), (2,2), activation="elu")(x)
x = tf.keras.layers.Conv2D(48, (5,5), (2,2), activation="elu")(x)
x = tf.keras.layers.Conv2D(64, (3,3), activation="elu")(x)
x = tf.keras.layers.Dropout(0.2)(x)
x = tf.keras.layers.Conv2D(64, (3,3), activation="elu")(x)
x = tf.keras.layers.Flatten()(x)
x = tf.keras.layers.Dropout(0.2)(x)
x = tf.keras.layers.Dense(100, activation='elu')(x)
x = tf.keras.layers.Dense(50, activation='elu')(x)
x = tf.keras.layers.Dense(10, activation='elu')(x)
outputs = tf.keras.layers.Dense(1)(x)
model = tf.keras.Model(inputs=inputs, outputs=outputs)
print(model.summary())
optimizer = Adam(learning_rate=1e-3)
model.compile(loss='mse', optimizer=optimizer)
img_gen = image_data_generator(X_train, y_train, batch_size=1)
X = next(img_gen)[0]
print(X.shape)
print(model(X).shape)
# saves the model weights after each epoch if the validation loss decreased
checkpoint_callback = keras.callbacks.ModelCheckpoint(filepath=os.path.join(model_output_dir,'lane_navigation_new_nvmodel_big2'), verbose=1, save_best_only=True) # revise
with tf.device('/device:GPU:0'):
history = model.fit(image_data_generator( X_train, y_train, batch_size=BATCH_SIZE), # 重要！batch_size大了好训练，1个batch_size训不出东西！
steps_per_epoch=10, # 300
epochs=300,
validation_data = image_data_generator( X_valid, y_valid, batch_size=BATCH_SIZE),
validation_steps=10, # 200
verbose=1,
shuffle=1,
callbacks=[checkpoint_callback])
Related
I am using a Keras CNN for handwritten digit recognition. I downloaded dataset from Kaggle.The preprocessing is :
train = pd.read_csv("./input/train.csv")
test = pd.read_csv("./input/test.csv")
Y_train = train["label"]
X_train = train.drop(labels = ["label"],axis = 1)
X_train = X_train / 255.0
test = test / 255.0
X_train = X_train.values.reshape(-1,28,28,1)
test = test.values.reshape(-1,28,28,1)
Y_train = to_categorical(Y_train, num_classes = 10)
random_seed = 2
X_train, X_val, Y_train, Y_val = train_test_split(X_train, Y_train, test_size = 0.1, random_state=random_seed)
My model looks like:
model = Sequential()
model.add(Conv2D(filters = 32, kernel_size = (5,5),padding = 'Same',
activation ='relu', input_shape = (28,28,1)))
model.add(Conv2D(filters = 32, kernel_size = (5,5),padding = 'Same',
activation ='relu'))
model.add(MaxPool2D(pool_size=(2,2)))
model.add(Dropout(0.25))
model.add(Conv2D(filters = 64, kernel_size = (3,3),padding = 'Same',
activation ='relu'))
model.add(Conv2D(filters = 64, kernel_size = (3,3),padding = 'Same',
activation ='relu'))
model.add(MaxPool2D(pool_size=(2,2), strides=(2,2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(256, activation = "relu"))
model.add(Dropout(0.5))
model.add(Dense(10, activation = "softmax"))
optimizer = RMSprop(lr=0.001, rho=0.9, epsilon=1e-08, decay=0.0)
model.compile(optimizer = optimizer , loss = "categorical_crossentropy", metrics=["accuracy"])
learning_rate_reduction = ReduceLROnPlateau(monitor='val_loss',
patience=3,
verbose=1,
factor=0.5,
min_lr=0.00001)
epochs = 1
batch_size = 86
datagen = ImageDataGenerator(
featurewise_center=False, # set input mean to 0 over the dataset
samplewise_center=False, # set each sample mean to 0
featurewise_std_normalization=False, # divide inputs by std of the dataset
samplewise_std_normalization=False, # divide each input by its std
zca_whitening=False, # apply ZCA whitening
rotation_range=10, # randomly rotate images in the range (degrees, 0 to 180)
zoom_range = 0.1, # Randomly zoom image
width_shift_range=0.1, # randomly shift images horizontally (fraction of total width)
height_shift_range=0.1, # randomly shift images vertically (fraction of total height)
horizontal_flip=False, # randomly flip images
vertical_flip=False) # randomly flip images
datagen.fit(X_train)
history = model.fit(datagen.flow(X_train,Y_train, batch_size=batch_size),
epochs = epochs,
validation_data = (X_val,Y_val),
verbose = 2,
steps_per_epoch=X_train.shape[0] // batch_size,
callbacks=[learning_rate_reduction])
Predicting my input:
def predict(image):
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
image = cv2.resize(image, (28, 28))
image = image.astype('float32')
image = image.reshape(1, 28, 28, 1)
image /= 255
model = load_model('./model.h5')
pred = model.predict(image, batch_size=1)
print("Predicted Number: ", pred.argmax())
predict(cv2.imread('./testImages/1.png'))
What am I doing wrong?
Desired result is the digit that is provided as an input image, instead I get the same output (i.e. digit 8) for every input.
The callback is saving checkpoint files, but not the SavedModel model.pb file. Additionally, when I load the model from the checkpoints it does not reload 'val_loss' which I'm conditioning "save_best_model" on.
I tried using a model.save() only on the best iteration but was having trouble with getting that to work correctly and it would be more convenient to use the ModelCheckpoint callback.
Here is the relevant code
LOSS = tf.keras.losses.MeanSquaredError(),
#multi output 3 categories from 0 to 1
model = ImgToClassSimpleContinuous(img_height, img_width)
checkpoint_filename = "../chkpts/ImgToClassSimpleContinuous/checkpoint_dir"
model.load_weights(checkpoint_filename)
cp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_filename,
verbose=1,mode='min', monitor="val_loss", save_best_only=True, save_weights_only=False)
model.compile(
optimizer='adam',
loss = [LOSS, LOSS, LOSS],
metrics=['mse'])
model.fit(
dataset_to_use,
validation_data = dataset_validation_batched,
# validation_steps=50,
epochs=MAX_EPOCHS,
batch_size=BATCH_SIZE,
callbacks=[cp_callback]
)
class ImgToClassSimpleContinuous(Model):
'''
pair with loss = categorical_crossentropy
'''
in_types = [DataType.d]
out_types = [DataType.tlc, DataType.tls, DataType.tll]
def __init__(self, img_height, img_width, *args, **kwargs):
super().__init__(ImgToClassSimple, *args, **kwargs)
initializer = 'he_normal'
input_shape = (img_height, img_width, 1)
inputs = tf.keras.Input(shape=input_shape)
flat_pix = layers.Flatten()(inputs)
x = layers.Conv2D(8, 3, padding='same', kernel_initializer=initializer)(inputs)
x = layers.PReLU()(x)
x = layers.Conv2D(8, 3, padding='same', kernel_initializer=initializer)(x)
x = layers.PReLU()(x)
x = layers.MaxPooling2D(pool_size=(2, 2))(x)
x = layers.BatchNormalization()(x)
x = layers.Conv2D(16, 3, padding='same', kernel_initializer=initializer)(x)
x = layers.PReLU()(x)
x = layers.Conv2D(16, 3, padding='same', kernel_initializer=initializer)(x)
x = layers.PReLU()(x)
x = layers.MaxPooling2D(pool_size=(2, 2))(x)
x = layers.BatchNormalization()(x)
t = layers.Conv2D(32, 3, padding='same', kernel_initializer=initializer)(x)
t = layers.PReLU()(t)
t = layers.Conv2D(32, 3, padding='same', kernel_initializer=initializer)(t)
t = layers.PReLU()(t)
t = layers.MaxPooling2D(pool_size=(2, 2))(t)
t = layers.BatchNormalization()(t)
t = tf.keras.layers.GlobalAveragePooling2D()(t)
t = layers.Flatten()(t)
s = layers.Conv2D(32, 3, padding='same', kernel_initializer=initializer)(x)
s = layers.PReLU()(s)
s = layers.Conv2D(32, 3, padding='same', kernel_initializer=initializer)(s)
s = layers.PReLU()(s)
s = layers.MaxPooling2D(pool_size=(2, 2))(s)
s = layers.BatchNormalization()(s)
s = tf.keras.layers.GlobalAveragePooling2D()(s)
s = layers.Flatten()(s)
l = layers.Conv2D(32, 3, padding='same', kernel_initializer=initializer)(x)
l = layers.PReLU()(l)
l = layers.Conv2D(32, 3, padding='same', kernel_initializer=initializer)(l)
l = layers.PReLU()(l)
l = layers.MaxPooling2D(pool_size=(2, 2))(l)
l = layers.BatchNormalization()(l)
l = tf.keras.layers.GlobalAveragePooling2D()(l)
l = layers.Flatten()(l)
t = layers.Dense(1, activation='sigmoid')(t)
s = layers.Dense(1, activation='sigmoid')(s)
l = layers.Dense(1, activation='sigmoid')(l)
# A Dense classifier with a single unit (binary classification)
self.model = tf.keras.Model(inputs, [t, s, l])
tf.keras.utils.plot_model(self.model, to_file="...", show_shapes=True)
def call(self, x):
return self.model(x)
I am working on a hand character recognition model. I created a CNN+BiLSTM+CTC Loss model. But getting error when I run model.fit(). Please help me fix this error.
My Model
# input with shape of height=32 and width=128
inputs = Input(shape=(32,128,1))
# convolution layer with kernel size (3,3)
conv_1 = Conv2D(64, (3,3), activation = 'relu', padding='same')(inputs)
# poolig layer with kernel size (2,2)
pool_1 = MaxPooling2D(pool_size=(2, 2), strides=2)(conv_1)
conv_2 = Conv2D(128, (3,3), activation = 'relu', padding='same')(pool_1)
pool_2 = MaxPooling2D(pool_size=(2, 2), strides=2)(conv_2)
conv_3 = Conv2D(256, (3,3), activation = 'relu', padding='same')(pool_2)
conv_4 = Conv2D(256, (3,3), activation = 'relu', padding='same')(conv_3)
# poolig layer with kernel size (2,1)
pool_4 = MaxPooling2D(pool_size=(2, 1))(conv_4)
conv_5 = Conv2D(512, (3,3), activation = 'relu', padding='same')(pool_4)
# Batch normalization layer
batch_norm_5 = BatchNormalization()(conv_5)
conv_6 = Conv2D(512, (3,3), activation = 'relu', padding='same')(batch_norm_5)
batch_norm_6 = BatchNormalization()(conv_6)
pool_6 = MaxPooling2D(pool_size=(2, 1))(batch_norm_6)
conv_7 = Conv2D(512, (2,2), activation = 'relu')(pool_6)
squeezed = Lambda(lambda x: K.squeeze(x, 1))(conv_7)
# bidirectional LSTM layers with units=128
blstm_1 = Bidirectional(LSTM(128, return_sequences=True, dropout = 0.2))(squeezed)
blstm_2 = Bidirectional(LSTM(128, return_sequences=True, dropout = 0.2))(blstm_1)
outputs = Dense(len(char_dict)+1, activation = 'softmax')(blstm_2)
act_model = Model(inputs, outputs)
Define a CTC loss model that takes the outputs of previous model as inputs
labels = Input(name='the_labels', shape=[max_length], dtype='float32')
input_length = Input(name='input_length', shape=[1], dtype='int64')
label_length = Input(name='label_length', shape=[1], dtype='int64')
def ctc_lambda_func(args):
y_pred, labels, input_length, label_length = args
return K.ctc_batch_cost(labels, y_pred, input_length, label_length)
loss_out = Lambda(ctc_lambda_func, output_shape=(1,), name='ctc')([outputs, labels, input_length,
label_length])
model = Model(inputs=[inputs, labels, input_length, label_length], outputs=loss_out)
model.compile(loss={'ctc': lambda y_true, y_pred: y_pred}, optimizer = 'adam')
model.fit(x=[input_array,
output_array,
train_input_length,
train_label_length],
y=np.zeros(input_array.shape[0]),
batch_size=256,
epochs = 100,
validation_data = ([test_input_array, test_output_array, valid_input_length,
valid_label_length], [np.zeros(test_input_array.shape[0])]),
verbose = 1,
callbacks = callbacks_list)
The error I am getting is
ValueError: Shape (None, 17) must have rank 1
I want to use TensorFlow Keras(v2.2) model fit in mnist with multiple outputs and losses, but it failed.
My costume model will return a list [logits, embedding]. logits is 2D tensor [batch , 10] and embedding is also 2D tensor [batch, 64].
class MyModel(tf.keras.Model):
def __init__(self):
super(MyModel, self).__init__()
self.reshape = tf.keras.layers.Reshape((28, 28, 1))
self.conv2D1 = tf.keras.layers.Conv2D(filters=8, kernel_size=(3,3), strides=(1, 1), padding='same', activation='relu')
self.maxPool1 = tf.keras.layers.MaxPooling2D(pool_size=(2, 2), strides=(2, 2), padding="same")
self.conv2D2 = tf.keras.layers.Conv2D(filters=8, kernel_size=(3,3), strides=(1, 1), padding='same', activation='relu')
self.maxPool2 = tf.keras.layers.MaxPooling2D(pool_size=2)
self.flatten = tf.keras.layers.Flatten(data_format="channels_last")
self.dropout = tf.keras.layers.Dropout(tf.compat.v1.placeholder_with_default(0.25, shape=[], name="dropout"))
self.dense1 = tf.keras.layers.Dense(64, activation=None)
self.dense2 = tf.keras.layers.Dense(10, activation=None)
def call(self, inputs, training):
x = self.reshape(inputs)
x = self.conv2D1(x)
x = self.maxPool1(x)
if training:
x = self.dropout(x)
x = self.conv2D2(x)
x = self.maxPool2(x)
if training:
x = self.dropout(x)
x = self.flatten(x)
x = self.dense1(x)
embedding = tf.math.l2_normalize(x, axis=1)
logits = self.dense2(embedding)
return [logits, embedding]
loss_0 is normal cross_entropy
def loss_0(y_true, y_pred):
loss_0 = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(labels=y_true, logits=y_pred[0]))
loss_1 is triplet_semihard_loss
def loss_1(y_true, y_pred):
loss_1 = tfa.losses.triplet_semihard_loss(y_true=y_true, y_pred=y_pred[1], distance_metric="L2")
return loss_1
When I use model fit, I can only get logits tensor in each loss. I can't get embedding tensor. y_pred[0] and y_pred[1] is not work. Any suggestion?
model = MyModel()
model.compile(optimizer=tf.keras.optimizers.Adam(lr=1e-3), loss=[loss_0, loss_1], loss_weights=[0.1, 0.1])
history = model.fit(train_dataset, epochs=5)
This should be a regression problem.
I would like the Neural Network to be able to estimate the length of a line, in pixel, from an image, like this 3 images, each image is 200 x 200 pcs:
a)b)c)
Training image of 6000 pcs, and validation image of 1000 pcs.
The labels are the distance in pixel:
a) 1.205404496424333018e+02
b) 1.188780888137086436e+02
c) 1.110180165558725918e+02
Here is my training code:
img_size = 200
def preprocess_image(image):
image = tf.image.decode_jpeg(image, channels=3)
image = tf.image.resize(image, [img_size, img_size])
image /= 255.0 # normalize to [0,1] range
return image
def load_and_preprocess_image(path):
image = tf.read_file(path)
return preprocess_image(image)
AUTOTUNE = tf.data.experimental.AUTOTUNE
BATCH_SIZE = 16
train_labels = np.loadtxt("train_labels.txt")
val_labels = np.loadtxt("test_labels.txt")
train_images = sorted(glob.glob("train_img/img_*.jpg"))
val_images = sorted(glob.glob("test_img/img_*.jpg"))
steps_per_epoch_count=tf.ceil(len(train_images)/BATCH_SIZE)
train_path_ds = tf.data.Dataset.from_tensor_slices(train_images)
val_path_ds = tf.data.Dataset.from_tensor_slices(val_images)
train_image_ds = train_path_ds.map(load_and_preprocess_image,
num_parallel_calls = AUTOTUNE)
train_label_ds =
tf.data.Dataset.from_tensor_slices(tf.cast(train_labels, tf.float32))
train_image_label_ds = tf.data.Dataset.zip((train_image_ds,
train_label_ds))
val_image_ds = val_path_ds.map(load_and_preprocess_image,
num_parallel_calls = AUTOTUNE)
val_label_ds = tf.data.Dataset.from_tensor_slices(tf.cast(val_labels, tf.float32))
val_image_label_ds = tf.data.Dataset.zip((val_image_ds, val_label_ds))
model = tf.keras.models.Sequential([
tf.keras.layers.Convolution2D(16,3,3, input_shape=(img_size,
img_size, 3), activation = 'relu'),
tf.keras.layers.MaxPooling2D(pool_size=(2,2)),
tf.keras.layers.Convolution2D(32,3,3, activation = 'relu'),
tf.keras.layers.MaxPooling2D(pool_size=(2,2)),
# tf.keras.layers.Convolution2D(64,3,3, activation = 'relu'),
# tf.keras.layers.MaxPooling2D(pool_size=(2,2)),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(400, activation=tf.nn.relu),
tf.keras.layers.Dropout(0.2),
tf.keras.layers.Dense(200, activation=tf.nn.relu),
tf.keras.layers.Dropout(0.1),
tf.keras.layers.Dense(100, activation=tf.nn.relu),
tf.keras.layers.Dropout(0.05),
tf.keras.layers.Dense(1, activation=tf.nn.relu)
])
model.compile(optimizer=tf.keras.optimizers.RMSprop(0.01),
loss = "mean_squared_error",
metrics = ["mean_absolute_error", "mean_squared_error"]
)
train_ds = train_image_label_ds.apply(tf.data.experimental.shuffle_and_repeat(buffer_size=len(train_images)))
train_ds = train_ds.batch(BATCH_SIZE)
train_ds = train_ds.prefetch(buffer_size=AUTOTUNE)
val_ds = val_image_label_ds.apply(
tf.data.experimental.shuffle_and_repeat(buffer_size=len(val_images)))
val_ds = val_ds.batch(BATCH_SIZE)
val_ds = val_ds.prefetch(buffer_size=AUTOTUNE)
history = model.fit(
train_ds,
epochs = 80,
validation_data = val_ds,
steps_per_epoch = 374,
validation_steps = 62
)
However, this is the train vs eval mean_squared_error plot:
Question:
Why is the validation loss not stable?
The average Mean Squared Error is about 400 in training, which seems too high. What modification I can do to improve the estimation?
EDIT:
This is my latest model:
Learning rate = 0.01
Batch size = 16
model = tf.keras.models.Sequential([
tf.keras.layers.Convolution2D(16,3,3, input_shape=(img_size, img_size, 3), activation = 'relu'),
tf.keras.layers.MaxPooling2D(pool_size=(2,2)),
tf.keras.layers.Convolution2D(32,3,3, activation = 'relu'),
tf.keras.layers.MaxPooling2D(pool_size=(2,2)),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(2, activation=tf.nn.relu),
tf.keras.layers.Dropout(0.5),
tf.keras.layers.Dense(2, activation=tf.nn.relu), #, kernel_regularizer = tf.keras.regularizers.l2(0.001)
tf.keras.layers.Dropout(0.5),
tf.keras.layers.Dense(2, activation=tf.nn.relu), #, kernel_regularizer = tf.keras.regularizers.l2(0.001)
tf.keras.layers.Dropout(0.5),
tf.keras.layers.Dense(2, activation=tf.nn.relu), #, kernel_regularizer = tf.keras.regularizers.l2(0.001)
tf.keras.layers.Dense(1, activation="linear")
])
The output looks like this:
As you can see, the train and validation loss is almost identical. The mse loss are both stabilized around 2393, which square root to 48.91 pixel error, quite high.
What advice to lower it further? Is it normal?