Tensorflow Convolution SoftmaxCrossEntropyWithLogits logits 和标签必须大小相同：logits_size=[640,2] labels_size=[10,2]
Tensorflow Convolution SoftmaxCrossEntropyWithLogits logits and labels must be same size: logits_size=[640,2] labels_size=[10,2]
我对每个卷积神经元层的矩阵计算有问题......
由此我得到：
InvalidArgumentError（回溯见上）：logits 和标签必须大小相同：logits_size=[640,2] labels_size=[10,2]
谁能给我指点适合初学者的资源并提供详细说明？
谢谢
来自 https://github.com/martin-gorner/tensorflow-mnist-tutorial/blob/master/mnist_3.0_convolutional.py
的源代码参考
import os
import tensorflow as tf
from tensorflow.python.framework import ops
from tensorflow.python.framework import dtypes
import numpy as np
import glob
import fnmatch
import matplotlib.pyplot as plt
from PIL import Image
import random
import threading
import math
tf.set_random_seed(0)
def convertToOneHot(vector, num_classes=None):
assert isinstance(vector, np.ndarray)
assert len(vector) > 0
if num_classes is None:
num_classes = np.max(vector)+1
else:
assert num_classes > 0
assert num_classes >= np.max(vector)
result = np.zeros(shape=(len(vector), num_classes))
result[np.arange(len(vector)), vector] = 1
return result.astype(np.float32)
def make_labels(filenames):
n = len(filenames)
#y = np.zeros((n,2), dtype = np.int32)
#y = np.zeros(shape=[n], dtype = np.float32)
label_y = np.zeros((n,2), dtype = np.float32)
counter = 0
dog = 0
cat = 0
for i in range(n):
# If 'dog' string is in file name assign '1'
if fnmatch.fnmatch(filenames[i], '*dog*'):
label_y[i,0] = 1
#label_y[i] = 1
dog += 1
else:
label_y[i,1] = 1
#label_y[i] = 0
cat += 1
print("Dog: " , dog , " Cat: " , cat)
return label_y
def make_test_labels(filenames):
n = len(filenames)
test_label_y = np.zeros([n], dtype = np.int32)
for i in range(n):
test_label_y[i] = random.randrange(0,2)
one_hot = convertToOneHot(test_label_y)
return one_hot
train_path = "./data/train/*.jpg"
test_path = "./data/test1/*.jpg"
#Training Dataset
train_files = tf.gfile.Glob(train_path)
train_image_labels = make_labels(train_files)
train_filename_queue = tf.train.string_input_producer(train_files, shuffle=False)
train_image_reader = tf.WholeFileReader()
train_image_filename, train_image_file = train_image_reader.read(train_filename_queue)
train_image_file = tf.image.decode_jpeg(train_image_file, 1)
train_image_file = tf.image.resize_images(train_image_file, [224, 224])
train_image_file.set_shape((224, 224, 1))
train_image_file = tf.squeeze(train_image_file)
#Test or Eval Dataset
test_files = tf.gfile.Glob(test_path)
test_image_labels = make_test_labels(test_files)
test_filename_queue = tf.train.string_input_producer(test_files, shuffle=False)
test_image_reader = tf.WholeFileReader()
test_image_filename, test_image_file = test_image_reader.read(test_filename_queue)
test_image_file = tf.image.decode_jpeg(test_image_file, 1)
test_image_file = tf.image.resize_images(test_image_file, [224, 224])
test_image_file.set_shape((224, 224, 1))
test_image_file = tf.squeeze(test_image_file)
train_batch_size = 10
test_batch_size = 2
num_preprocess_threads = 1
min_queue_examples = 256
X = tf.placeholder(tf.float32, [None, 224, 224, 1])
Y_ = tf.placeholder(tf.float32, [None, 2])
lr = tf.placeholder(tf.float32)
pkeep = tf.placeholder(tf.float32)
# three convolutional layers with their channel counts, and a
# fully connected layer (tha last layer has 2 softmax neurons)
K = 4 # first convolutional layer output depth
L = 8 # second convolutional layer output depth
M = 12 # third convolutional layer
N = 200 # fully connected layer
W1 = tf.Variable(tf.truncated_normal([5, 5, 1, K], stddev=0.1)) # 5x5 patch, 1 input channel, K output channels
print "W1: " , W1.get_shape()
B1 = tf.Variable(tf.ones([K])/10)
print "B1: " , B1.get_shape()
W2 = tf.Variable(tf.truncated_normal([5, 5, K, L], stddev=0.1))
print "W2: " , W2.get_shape()
B2 = tf.Variable(tf.ones([L])/10)
print "B2: " , B2.get_shape()
W3 = tf.Variable(tf.truncated_normal([4, 4, L, M], stddev=0.1))
print "W3: " , W3.get_shape()
B3 = tf.Variable(tf.ones([M])/10)
print "B3: " , B3.get_shape()
W4 = tf.Variable(tf.truncated_normal([7 * 7 * M, N], stddev=0.1))
print "W4: " , W4.get_shape()
B4 = tf.Variable(tf.ones([N])/10)
print "B4: " , B4.get_shape()
W5 = tf.Variable(tf.truncated_normal([N, 2], stddev=0.1))
print "W5: " , W5.get_shape()
B5 = tf.Variable(tf.ones([2])/10)
print "B5: " , B5.get_shape()
# The model
stride = 1 # output is 28x28
Y1 = tf.nn.relu(tf.nn.conv2d(X, W1, strides=[1, stride, stride, 1], padding='SAME') + B1)
print "Y1: " , Y1.get_shape()
stride = 2 # output is 14x14
Y2 = tf.nn.relu(tf.nn.conv2d(Y1, W2, strides=[1, stride, stride, 1], padding='SAME') + B2)
print "Y2: " , Y2.get_shape()
stride = 2 # output is 7x7
Y3 = tf.nn.relu(tf.nn.conv2d(Y2, W3, strides=[1, stride, stride, 1], padding='SAME') + B3)
print "Y3: " , Y3.get_shape()
# reshape the output from the third convolution for the fully connected layer
#YY = tf.reshape(Y3, shape=[-1, 7 * 7 * M])
YY = tf.reshape(Y3, shape=[-1, 7 * 7 * M])
print "YY: " , YY.get_shape()
Y4 = tf.nn.relu(tf.matmul(YY, W4) + B4)
print "Y4: " , Y4.get_shape()
Ylogits = tf.matmul(Y4, W5) + B5
print "Ylogits: " , Ylogits.get_shape()
Y = tf.nn.softmax(Ylogits)
# cross-entropy loss function (= -sum(Y_i * log(Yi)) ), normalised for batches of 10 images
# TensorFlow provides the softmax_cross_entropy_with_logits function to avoid numerical stability
# problems with log(0) which is NaN
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=Ylogits, labels=Y_)
cross_entropy = tf.reduce_mean(cross_entropy) * 10
# accuracy of the trained model, between 0 (worst) and 1 (best)
correct_prediction = tf.equal(tf.argmax(Y, 1), tf.argmax(Y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
allweights = tf.concat(0, [tf.reshape(W1, [-1]), tf.reshape(W2, [-1]), tf.reshape(W3, [-1]), tf.reshape(W4, [-1]), tf.reshape(W5, [-1])])
allbiases = tf.concat(0, [tf.reshape(B1, [-1]), tf.reshape(B2, [-1]), tf.reshape(B3, [-1]), tf.reshape(B4, [-1]), tf.reshape(B5, [-1])])
# training step, the learning rate is a placeholder
train_step = tf.train.AdamOptimizer(lr).minimize(cross_entropy)
train_images = tf.train.batch([train_image_file], batch_size=train_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * train_batch_size, allow_smaller_final_batch=True)
test_images = tf.train.batch([test_image_file], batch_size=test_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * test_batch_size, allow_smaller_final_batch=True)
train_labels = tf.train.batch([train_image_labels], batch_size=train_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * train_batch_size, enqueue_many=True, allow_smaller_final_batch=True)
test_labels = tf.train.batch([test_image_labels], batch_size=test_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * test_batch_size, enqueue_many=True, allow_smaller_final_batch=True)
init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)
def training_step(i, update_test_data, update_train_data):
train_images_batch = train_images.eval(session=sess)
train_images_batch = np.expand_dims(train_images_batch, axis=(3))
train_labels_batch = train_labels.eval(session=sess)
test_images_batch = test_images.eval(session=sess)
test_images_batch = np.expand_dims(test_images_batch, axis=(3))
test_labels_batch = test_labels.eval(session=sess)
# learning rate decay
max_learning_rate = 0.003
min_learning_rate = 0.0001
decay_speed = 2000.0
learning_rate = min_learning_rate + (max_learning_rate - min_learning_rate) * math.exp(-i/decay_speed)
if update_train_data:
a, c, w, b = sess.run([accuracy, cross_entropy, allweights, allbiases], {X: train_images_batch, Y_: train_labels_batch})
print(str(i) + ": accuracy:" + str(a) + " loss: " + str(c) + " (lr:" + str(learning_rate) + ")")
if update_test_data:
a, c = sess.run([accuracy, cross_entropy], {X: test_images_batch, Y_: test_labels_batch})
print(str(i) + ": ********* epoch " + " ********* test accuracy:" + str(a) + " test loss: " + str(c))
# the backpropagation training step
sess.run(train_step, {X: batch_X, Y_: batch_Y, lr: learning_rate})
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(coord=coord, sess=sess)
for i in range(10000+1):
training_step(i, i % 100 == 0, i % 20 == 0)
coord.request_stop()
coord.join(threads)
结果：
('Dog: ', 12500, ' Cat: ', 12500)
W1: (5, 5, 1, 4)
B1: (4,)
W2: (5, 5, 4, 8)
B2: (8,)
W3: (4, 4, 8, 12)
B3: (12,)
W4: (588, 200)
B4: (200,)
W5: (200, 2)
B5: (2,)
Y1: (?, 224, 224, 4)
Y2: (?, 112, 112, 8)
Y3: (?, 56, 56, 12)
YY: (?, 588)
Y4: (?, 200)
Ylogits: (?, 2)
Traceback (most recent call last):
File "convolutional.py", line 306, in <module>
training_step(i, i % 100 == 0, i % 20 == 0)
File "convolutional.py", line 288, in training_step
a, c, w, b = sess.run([accuracy, cross_entropy, allweights, allbiases], {X: train_images_batch, Y_: train_labels_batch})
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 766, in run
run_metadata_ptr)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 964, in _run
feed_dict_string, options, run_metadata)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1014, in _do_run
target_list, options, run_metadata)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1034, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.InvalidArgumentError: logits and labels must be same size: logits_size=[640,2] labels_size=[10,2]
[[Node: SoftmaxCrossEntropyWithLogits = SoftmaxCrossEntropyWithLogits[T=DT_FLOAT, _device="/job:localhost/replica:0/task:0/cpu:0"](Reshape_1, Reshape_2)]]
Caused by op u'SoftmaxCrossEntropyWithLogits', defined at:
File "convolutional.py", line 229, in <module>
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=Ylogits, labels=Y_)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/ops/nn_ops.py", line 1449, in softmax_cross_entropy_with_logits
precise_logits, labels, name=name)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/ops/gen_nn_ops.py", line 2265, in _softmax_cross_entropy_with_logits
features=features, labels=labels, name=name)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 759, in apply_op
op_def=op_def)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2240, in create_op
original_op=self._default_original_op, op_def=op_def)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1128, in __init__
self._traceback = _extract_stack()
InvalidArgumentError (see above for traceback): logits and labels must be same size: logits_size=[640,2] labels_size=[10,2]
[[Node: SoftmaxCrossEntropyWithLogits = SoftmaxCrossEntropyWithLogits[T=DT_FLOAT, _device="/job:localhost/replica:0/task:0/cpu:0"](Reshape_1, Reshape_2)]]
设法修复它：
import os
import tensorflow as tf
from tensorflow.python.framework import ops
from tensorflow.python.framework import dtypes
import numpy as np
import glob
import fnmatch
import matplotlib.pyplot as plt
from PIL import Image
import random
import threading
import math
tf.set_random_seed(0)
def convertToOneHot(vector, num_classes=None):
assert isinstance(vector, np.ndarray)
assert len(vector) > 0
if num_classes is None:
num_classes = np.max(vector)+1
else:
assert num_classes > 0
assert num_classes >= np.max(vector)
result = np.zeros(shape=(len(vector), num_classes))
result[np.arange(len(vector)), vector] = 1
return result.astype(np.float32)
def make_labels(filenames):
n = len(filenames)
#y = np.zeros((n,2), dtype = np.int32)
#y = np.zeros(shape=[n], dtype = np.float32)
label_y = np.zeros((n,2), dtype = np.float32)
counter = 0
dog = 0
cat = 0
for i in range(n):
# If 'dog' string is in file name assign '1'
if fnmatch.fnmatch(filenames[i], '*dog*'):
label_y[i,0] = 1
#label_y[i] = 1
dog += 1
else:
label_y[i,1] = 1
#label_y[i] = 0
cat += 1
print("Dog: " , dog , " Cat: " , cat)
return label_y
def make_test_labels(filenames):
n = len(filenames)
test_label_y = np.zeros([n], dtype = np.int32)
for i in range(n):
test_label_y[i] = random.randrange(0,2)
one_hot = convertToOneHot(test_label_y)
return one_hot
train_path = "./data/train/*.jpg"
test_path = "./data/test1/*.jpg"
#Training Dataset
train_files = tf.gfile.Glob(train_path)
train_image_labels = make_labels(train_files)
train_filename_queue = tf.train.string_input_producer(train_files, shuffle=False)
train_image_reader = tf.WholeFileReader()
train_image_filename, train_image_file = train_image_reader.read(train_filename_queue)
train_image_file = tf.image.decode_jpeg(train_image_file, 1)
train_image_file = tf.image.resize_images(train_image_file, [224, 224])
train_image_file.set_shape((224, 224, 1))
train_image_file = tf.squeeze(train_image_file)
#Test or Eval Dataset
test_files = tf.gfile.Glob(test_path)
test_image_labels = make_test_labels(test_files)
test_filename_queue = tf.train.string_input_producer(test_files, shuffle=False)
test_image_reader = tf.WholeFileReader()
test_image_filename, test_image_file = test_image_reader.read(test_filename_queue)
test_image_file = tf.image.decode_jpeg(test_image_file, 1)
test_image_file = tf.image.resize_images(test_image_file, [224, 224])
test_image_file.set_shape((224, 224, 1))
test_image_file = tf.squeeze(test_image_file)
train_batch_size = 10
test_batch_size = 2
num_preprocess_threads = 1
min_queue_examples = 256
X = tf.placeholder(tf.float32, [None, 224, 224, 1])
Y_ = tf.placeholder(tf.float32, [None, 2])
lr = tf.placeholder(tf.float32)
pkeep = tf.placeholder(tf.float32)
# three convolutional layers with their channel counts, and a
# fully connected layer (tha last layer has 2 softmax neurons)
K = 4 # first convolutional layer output depth
L = 8 # second convolutional layer output depth
M = 12 # third convolutional layer
N = 200 # fully connected layer
W1 = tf.Variable(tf.truncated_normal([5, 5, 1, K], stddev=0.1)) # 5x5 patch, 1 input channel, K output channels
print "W1: " , W1.get_shape()
B1 = tf.Variable(tf.ones([K])/10)
print "B1: " , B1.get_shape()
W2 = tf.Variable(tf.truncated_normal([5, 5, K, L], stddev=0.1))
print "W2: " , W2.get_shape()
B2 = tf.Variable(tf.ones([L])/10)
print "B2: " , B2.get_shape()
W3 = tf.Variable(tf.truncated_normal([4, 4, L, M], stddev=0.1))
print "W3: " , W3.get_shape()
B3 = tf.Variable(tf.ones([M])/10)
print "B3: " , B3.get_shape()
W4 = tf.Variable(tf.truncated_normal([7 * 7 * M, N], stddev=0.1))
print "W4: " , W4.get_shape()
B4 = tf.Variable(tf.ones([N])/10)
print "B4: " , B4.get_shape()
W5 = tf.Variable(tf.truncated_normal([N, 2], stddev=0.1))
print "W5: " , W5.get_shape()
B5 = tf.Variable(tf.ones([2])/10)
print "B5: " , B5.get_shape()
# The model
stride = 1
Y1 = tf.nn.relu(tf.nn.conv2d(X, W1, strides=[1, stride, stride, 1], padding='SAME') + B1)
print "Y1: " , Y1.get_shape()
stride = 6 #changed from 2 to 6
Y2 = tf.nn.relu(tf.nn.conv2d(Y1, W2, strides=[1, stride, stride, 1], padding='SAME') + B2)
print "Y2: " , Y2.get_shape()
stride = 6 #changed from 2 to 6
Y3 = tf.nn.relu(tf.nn.conv2d(Y2, W3, strides=[1, stride, stride, 1], padding='SAME') + B3)
print "Y3: " , Y3.get_shape()
# reshape the output from the third convolution for the fully connected layer
#YY = tf.reshape(Y3, shape=[-1, 7 * 7 * M])
YY = tf.reshape(Y3, shape=[-1, 7 * 7 * M])
print "YY: " , YY.get_shape()
Y4 = tf.nn.relu(tf.matmul(YY, W4) + B4)
print "Y4: " , Y4.get_shape()
Ylogits = tf.matmul(Y4, W5) + B5
print "Ylogits: " , Ylogits.get_shape()
Y = tf.nn.softmax(Ylogits)
# cross-entropy loss function (= -sum(Y_i * log(Yi)) ), normalised for batches of 10 images
# TensorFlow provides the softmax_cross_entropy_with_logits function to avoid numerical stability
# problems with log(0) which is NaN
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=Ylogits, labels=Y_)
cross_entropy = tf.reduce_mean(cross_entropy) * 10
# accuracy of the trained model, between 0 (worst) and 1 (best)
correct_prediction = tf.equal(tf.argmax(Y, 1), tf.argmax(Y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
allweights = tf.concat(0, [tf.reshape(W1, [-1]), tf.reshape(W2, [-1]), tf.reshape(W3, [-1]), tf.reshape(W4, [-1]), tf.reshape(W5, [-1])])
allbiases = tf.concat(0, [tf.reshape(B1, [-1]), tf.reshape(B2, [-1]), tf.reshape(B3, [-1]), tf.reshape(B4, [-1]), tf.reshape(B5, [-1])])
# training step, the learning rate is a placeholder
train_step = tf.train.AdamOptimizer(lr).minimize(cross_entropy)
train_images = tf.train.batch([train_image_file], batch_size=train_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * train_batch_size, allow_smaller_final_batch=True)
test_images = tf.train.batch([test_image_file], batch_size=test_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * test_batch_size, allow_smaller_final_batch=True)
train_labels = tf.train.batch([train_image_labels], batch_size=train_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * train_batch_size, enqueue_many=True, allow_smaller_final_batch=True)
test_labels = tf.train.batch([test_image_labels], batch_size=test_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * test_batch_size, enqueue_many=True, allow_smaller_final_batch=True)
init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)
def training_step(i, update_test_data, update_train_data):
train_images_batch = train_images.eval(session=sess)
train_images_batch = np.expand_dims(train_images_batch, axis=(3))
train_labels_batch = train_labels.eval(session=sess)
test_images_batch = test_images.eval(session=sess)
test_images_batch = np.expand_dims(test_images_batch, axis=(3))
test_labels_batch = test_labels.eval(session=sess)
# learning rate decay
max_learning_rate = 0.003
min_learning_rate = 0.0001
decay_speed = 2000.0
learning_rate = min_learning_rate + (max_learning_rate - min_learning_rate) * math.exp(-i/decay_speed)
if update_train_data:
a, c, w, b = sess.run([accuracy, cross_entropy, allweights, allbiases], {X: train_images_batch, Y_: train_labels_batch})
print(str(i) + ": accuracy:" + str(a) + " loss: " + str(c) + " (lr:" + str(learning_rate) + ")")
if update_test_data:
a, c = sess.run([accuracy, cross_entropy], {X: test_images_batch, Y_: test_labels_batch})
print(str(i) + ": ********* epoch " + " ********* test accuracy:" + str(a) + " test loss: " + str(c))
# the backpropagation training step
sess.run(train_step, {X: batch_X, Y_: batch_Y, lr: learning_rate})
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(coord=coord, sess=sess)
for i in range(10000+1):
training_step(i, i % 100 == 0, i % 20 == 0)
coord.request_stop()
coord.join(threads)
结果：
0: accuracy:0.6 loss: 11.1554 (lr:0.003)
0: ********* epoch ********* test accuracy:0.5 test loss: 13.0221
20: accuracy:0.7 loss: 5.93612 (lr:0.00297114451787)
40: accuracy:0.5 loss: 7.40466 (lr:0.00294257615259)
60: accuracy:0.4 loss: 6.7717 (lr:0.00291429204729)
80: accuracy:0.5 loss: 7.07674 (lr:0.00288628937354)
100: accuracy:0.6 loss: 7.0582 (lr:0.00285856533105)
100: ********* epoch ********* test accuracy:0.5 test loss: 7.8989
120: accuracy:0.3 loss: 6.95063 (lr:0.00283111714739)
140: accuracy:0.5 loss: 6.57362 (lr:0.00280394207773)
160: accuracy:0.4 loss: 7.82804 (lr:0.00277703740452)
180: accuracy:0.6 loss: 6.80486 (lr:0.00275040043729)
200: accuracy:0.6 loss: 6.98334 (lr:0.0027240285123)
200: ********* epoch ********* test accuracy:0.5 test loss: 6.71705
220: accuracy:0.9 loss: 4.97062 (lr:0.00269791899236)
240: accuracy:0.5 loss: 5.81752 (lr:0.00267206926648)
260: accuracy:0.6 loss: 6.53955 (lr:0.00264647674967)
280: accuracy:0.3 loss: 8.13235 (lr:0.00262113888266)
300: accuracy:0.5 loss: 6.55135 (lr:0.00259605313163)
300: ********* epoch ********* test accuracy:0.5 test loss: 6.69505
320: accuracy:0.7 loss: 6.20664 (lr:0.002571216988)
340: accuracy:0.6 loss: 6.23909 (lr:0.00254662796813)
360: accuracy:0.8 loss: 4.65571 (lr:0.00252228361309)
380: accuracy:0.5 loss: 8.66489 (lr:0.00249818148844)
400: accuracy:0.6 loss: 6.5182 (lr:0.00247431918393)
400: ********* epoch ********* test accuracy:0.0 test loss: 7.87964
420: accuracy:0.8 loss: 5.66837 (lr:0.00245069431331)
440: accuracy:0.5 loss: 6.87555 (lr:0.00242730451409)
460: accuracy:0.7 loss: 5.9571 (lr:0.00240414744726)
480: accuracy:0.6 loss: 6.63313 (lr:0.00238122079709)
500: accuracy:0.7 loss: 6.31318 (lr:0.00235852227091)
500: ********* epoch ********* test accuracy:0.0 test loss: 8.10029
520: accuracy:0.7 loss: 7.07258 (lr:0.00233604959883)
540: accuracy:0.5 loss: 7.05975 (lr:0.00231380053358)
560: accuracy:0.5 loss: 7.25364 (lr:0.00229177285022)
580: accuracy:0.6 loss: 6.73549 (lr:0.00226996434598)
600: accuracy:0.8 loss: 4.62922 (lr:0.00224837283998)
600: ********* epoch ********* test accuracy:0.5 test loss: 6.42755
620: accuracy:0.6 loss: 6.97526 (lr:0.00222699617305)
640: accuracy:0.7 loss: 5.79574 (lr:0.00220583220751)
660: accuracy:0.7 loss: 6.3626 (lr:0.00218487882695)
680: accuracy:0.8 loss: 7.46478 (lr:0.00216413393601)
700: accuracy:0.6 loss: 5.08642 (lr:0.00214359546018)
700: ********* epoch ********* test accuracy:0.5 test loss: 13.1772
720: accuracy:0.5 loss: 7.97948 (lr:0.00212326134561)
740: accuracy:0.5 loss: 7.21764 (lr:0.00210312955885)
760: accuracy:0.7 loss: 6.36856 (lr:0.00208319808672)
780: accuracy:0.7 loss: 5.55039 (lr:0.00206346493604)
800: accuracy:0.6 loss: 7.82716 (lr:0.0020439281335)
800: ********* epoch ********* test accuracy:0.0 test loss: 8.27366
820: accuracy:0.5 loss: 6.67392 (lr:0.0020245857254)
840: accuracy:0.6 loss: 6.52732 (lr:0.00200543577746)
860: accuracy:0.6 loss: 6.47456 (lr:0.0019864763747)
880: accuracy:0.8 loss: 5.34777 (lr:0.00196770562114)
900: accuracy:0.5 loss: 6.67296 (lr:0.0019491216397)
900: ********* epoch ********* test accuracy:0.0 test loss: 9.64283
920: accuracy:0.7 loss: 6.66137 (lr:0.00193072257197)
940: accuracy:0.9 loss: 5.7024 (lr:0.00191250657802)
960: accuracy:0.7 loss: 6.43189 (lr:0.00189447183624)
980: accuracy:0.7 loss: 5.07373 (lr:0.00187661654313)
1000: accuracy:0.7 loss: 6.38296 (lr:0.00185893891317)
1000: ********* epoch ********* test accuracy:0.5 test loss: 8.52038
1020: accuracy:0.8 loss: 9.55432 (lr:0.00184143717856)
1040: accuracy:0.6 loss: 7.09334 (lr:0.00182410958911)
1060: accuracy:0.4 loss: 7.11957 (lr:0.00180695441207)
1080: accuracy:0.6 loss: 7.0127 (lr:0.00178996993188)
1100: accuracy:0.4 loss: 7.69579 (lr:0.0017731544501)
1100: ********* epoch ********* test accuracy:0.5 test loss: 11.4886
1120: accuracy:0.7 loss: 6.882 (lr:0.00175650628516)
1140: accuracy:0.8 loss: 5.1929 (lr:0.00174002377223)
1160: accuracy:0.3 loss: 6.76535 (lr:0.00172370526304)
1180: accuracy:0.5 loss: 7.55367 (lr:0.00170754912573)
1200: accuracy:0.7 loss: 5.56166 (lr:0.00169155374467)
1200: ********* epoch ********* test accuracy:0.5 test loss: 7.25217
1220: accuracy:0.8 loss: 5.60191 (lr:0.00167571752032)
1240: accuracy:0.6 loss: 6.82252 (lr:0.00166003886902)
1260: accuracy:0.9 loss: 5.644 (lr:0.00164451622292)
1280: accuracy:0.8 loss: 5.7729 (lr:0.00162914802972)
1300: accuracy:0.7 loss: 5.87659 (lr:0.00161393275261)
1300: ********* epoch ********* test accuracy:0.5 test loss: 5.97413
1320: accuracy:0.5 loss: 7.00393 (lr:0.00159886887003)
1340: accuracy:0.5 loss: 6.74683 (lr:0.00158395487558)
1360: accuracy:0.5 loss: 6.34526 (lr:0.00156918927786)
1380: accuracy:0.5 loss: 7.20938 (lr:0.00155457060029)
1400: accuracy:0.5 loss: 6.90222 (lr:0.001540097381)
1400: ********* epoch ********* test accuracy:1.0 test loss: 5.80986
1420: accuracy:0.7 loss: 6.27892 (lr:0.00152576817264)
1440: accuracy:0.6 loss: 7.12203 (lr:0.00151158154228)
1460: accuracy:0.6 loss: 6.3593 (lr:0.00149753607126)
1480: accuracy:0.5 loss: 6.73234 (lr:0.00148363035501)
1500: accuracy:0.7 loss: 5.93877 (lr:0.00146986300295)
1500: ********* epoch ********* test accuracy:0.0 test loss: 13.2649
1520: accuracy:0.5 loss: 6.56755 (lr:0.00145623263833)
1540: accuracy:0.2 loss: 11.2914 (lr:0.0014427378981)
1560: accuracy:0.7 loss: 6.40623 (lr:0.00142937743279)
1580: accuracy:0.8 loss: 5.21643 (lr:0.00141614990632)
1600: accuracy:0.6 loss: 7.13112 (lr:0.00140305399594)
1600: ********* epoch ********* test accuracy:0.5 test loss: 7.00448
1620: accuracy:0.8 loss: 6.04726 (lr:0.00139008839205)
1640: accuracy:0.6 loss: 5.6949 (lr:0.00137725179807)
1660: accuracy:0.4 loss: 8.35133 (lr:0.00136454293033)
1680: accuracy:0.7 loss: 5.67702 (lr:0.00135196051794)
1700: accuracy:0.7 loss: 5.75319 (lr:0.00133950330265)
1700: ********* epoch ********* test accuracy:0.5 test loss: 4.69712
1720: accuracy:0.7 loss: 5.62581 (lr:0.00132717003872)
1740: accuracy:0.5 loss: 6.94164 (lr:0.00131495949282)
1760: accuracy:0.8 loss: 6.81993 (lr:0.00130287044388)
1780: accuracy:0.6 loss: 6.46852 (lr:0.00129090168298)
1800: accuracy:0.7 loss: 5.52436 (lr:0.00127905201325)
1800: ********* epoch ********* test accuracy:1.0 test loss: 5.64173
1820: accuracy:0.6 loss: 5.64919 (lr:0.0012673202497)
1840: accuracy:0.6 loss: 6.43817 (lr:0.00125570521915)
1860: accuracy:0.6 loss: 6.09095 (lr:0.00124420576008)
1880: accuracy:0.7 loss: 5.98002 (lr:0.00123282072254)
1900: accuracy:0.8 loss: 5.04342 (lr:0.00122154896802)
1900: ********* epoch ********* test accuracy:0.5 test loss: 5.80969
1920: accuracy:0.5 loss: 6.73896 (lr:0.00121038936933)
1940: accuracy:0.6 loss: 5.74824 (lr:0.0011993408105)
1960: accuracy:0.9 loss: 5.07425 (lr:0.00118840218667)
1980: accuracy:0.8 loss: 5.07815 (lr:0.00117757240396)
2000: accuracy:0.6 loss: 6.49868 (lr:0.0011668503794)
2000: ********* epoch ********* test accuracy:0.5 test loss: 5.17815
2020: accuracy:0.7 loss: 6.32198 (lr:0.00115623504076)
2040: accuracy:0.5 loss: 6.81228 (lr:0.0011457253265)
2060: accuracy:0.7 loss: 6.27763 (lr:0.00113532018565)
2080: accuracy:0.4 loss: 6.48217 (lr:0.00112501857768)
2100: accuracy:0.8 loss: 5.47317 (lr:0.00111481947242)
2100: ********* epoch ********* test accuracy:0.5 test loss: 6.97906
2120: accuracy:0.7 loss: 6.64062 (lr:0.00110472184996)
2140: accuracy:0.4 loss: 7.58298 (lr:0.00109472470051)
2160: accuracy:0.5 loss: 7.05855 (lr:0.00108482702437)
2180: accuracy:0.7 loss: 5.3101 (lr:0.00107502783175)
2200: accuracy:0.4 loss: 7.99556 (lr:0.00106532614272)
2200: ********* epoch ********* test accuracy:0.0 test loss: 30.0326
2220: accuracy:0.5 loss: 5.50862 (lr:0.00105572098712)
2240: accuracy:0.7 loss: 7.57991 (lr:0.00104621140441)
2260: accuracy:0.9 loss: 4.57821 (lr:0.00103679644362)
2280: accuracy:0.5 loss: 6.55349 (lr:0.00102747516327)
2300: accuracy:0.4 loss: 7.27755 (lr:0.0010182466312)
2300: ********* epoch ********* test accuracy:0.5 test loss: 8.20956
2320: accuracy:0.7 loss: 5.98563 (lr:0.00100910992456)
2340: accuracy:0.7 loss: 5.72076 (lr:0.00100006412967)
2360: accuracy:0.7 loss: 5.22523 (lr:0.000991108341943)
2380: accuracy:0.8 loss: 4.87689 (lr:0.000982241665793)
2400: accuracy:0.8 loss: 5.14703 (lr:0.000973463214545)
2400: ********* epoch ********* test accuracy:0.5 test loss: 6.0281
2420: accuracy:0.7 loss: 6.03927 (lr:0.000964772110347)
2440: accuracy:0.7 loss: 4.91816 (lr:0.00095616748408)
2460: accuracy:0.5 loss: 6.61033 (lr:0.000947648475274)
2480: accuracy:0.7 loss: 4.98853 (lr:0.000939214232023)
2500: accuracy:0.5 loss: 6.02745 (lr:0.000930863910895)
2500: ********* epoch ********* test accuracy:0.5 test loss: 6.64094
2520: accuracy:0.8 loss: 5.77956 (lr:0.000922596676849)
2540: accuracy:0.5 loss: 7.03326 (lr:0.000914411703157)
2560: accuracy:0.9 loss: 3.64503 (lr:0.000906308171314)
2580: accuracy:0.5 loss: 7.78428 (lr:0.00089828527096)
好像准确率不太好…………是不是要加层数？
还是哪里出了问题？
谢谢
我对每个卷积神经元层的矩阵计算有问题......
由此我得到：
InvalidArgumentError（回溯见上）：logits 和标签必须大小相同：logits_size=[640,2] labels_size=[10,2]
谁能给我指点适合初学者的资源并提供详细说明？
谢谢
来自 https://github.com/martin-gorner/tensorflow-mnist-tutorial/blob/master/mnist_3.0_convolutional.py
的源代码参考import os
import tensorflow as tf
from tensorflow.python.framework import ops
from tensorflow.python.framework import dtypes
import numpy as np
import glob
import fnmatch
import matplotlib.pyplot as plt
from PIL import Image
import random
import threading
import math
tf.set_random_seed(0)
def convertToOneHot(vector, num_classes=None):
assert isinstance(vector, np.ndarray)
assert len(vector) > 0
if num_classes is None:
num_classes = np.max(vector)+1
else:
assert num_classes > 0
assert num_classes >= np.max(vector)
result = np.zeros(shape=(len(vector), num_classes))
result[np.arange(len(vector)), vector] = 1
return result.astype(np.float32)
def make_labels(filenames):
n = len(filenames)
#y = np.zeros((n,2), dtype = np.int32)
#y = np.zeros(shape=[n], dtype = np.float32)
label_y = np.zeros((n,2), dtype = np.float32)
counter = 0
dog = 0
cat = 0
for i in range(n):
# If 'dog' string is in file name assign '1'
if fnmatch.fnmatch(filenames[i], '*dog*'):
label_y[i,0] = 1
#label_y[i] = 1
dog += 1
else:
label_y[i,1] = 1
#label_y[i] = 0
cat += 1
print("Dog: " , dog , " Cat: " , cat)
return label_y
def make_test_labels(filenames):
n = len(filenames)
test_label_y = np.zeros([n], dtype = np.int32)
for i in range(n):
test_label_y[i] = random.randrange(0,2)
one_hot = convertToOneHot(test_label_y)
return one_hot
train_path = "./data/train/*.jpg"
test_path = "./data/test1/*.jpg"
#Training Dataset
train_files = tf.gfile.Glob(train_path)
train_image_labels = make_labels(train_files)
train_filename_queue = tf.train.string_input_producer(train_files, shuffle=False)
train_image_reader = tf.WholeFileReader()
train_image_filename, train_image_file = train_image_reader.read(train_filename_queue)
train_image_file = tf.image.decode_jpeg(train_image_file, 1)
train_image_file = tf.image.resize_images(train_image_file, [224, 224])
train_image_file.set_shape((224, 224, 1))
train_image_file = tf.squeeze(train_image_file)
#Test or Eval Dataset
test_files = tf.gfile.Glob(test_path)
test_image_labels = make_test_labels(test_files)
test_filename_queue = tf.train.string_input_producer(test_files, shuffle=False)
test_image_reader = tf.WholeFileReader()
test_image_filename, test_image_file = test_image_reader.read(test_filename_queue)
test_image_file = tf.image.decode_jpeg(test_image_file, 1)
test_image_file = tf.image.resize_images(test_image_file, [224, 224])
test_image_file.set_shape((224, 224, 1))
test_image_file = tf.squeeze(test_image_file)
train_batch_size = 10
test_batch_size = 2
num_preprocess_threads = 1
min_queue_examples = 256
X = tf.placeholder(tf.float32, [None, 224, 224, 1])
Y_ = tf.placeholder(tf.float32, [None, 2])
lr = tf.placeholder(tf.float32)
pkeep = tf.placeholder(tf.float32)
# three convolutional layers with their channel counts, and a
# fully connected layer (tha last layer has 2 softmax neurons)
K = 4 # first convolutional layer output depth
L = 8 # second convolutional layer output depth
M = 12 # third convolutional layer
N = 200 # fully connected layer
W1 = tf.Variable(tf.truncated_normal([5, 5, 1, K], stddev=0.1)) # 5x5 patch, 1 input channel, K output channels
print "W1: " , W1.get_shape()
B1 = tf.Variable(tf.ones([K])/10)
print "B1: " , B1.get_shape()
W2 = tf.Variable(tf.truncated_normal([5, 5, K, L], stddev=0.1))
print "W2: " , W2.get_shape()
B2 = tf.Variable(tf.ones([L])/10)
print "B2: " , B2.get_shape()
W3 = tf.Variable(tf.truncated_normal([4, 4, L, M], stddev=0.1))
print "W3: " , W3.get_shape()
B3 = tf.Variable(tf.ones([M])/10)
print "B3: " , B3.get_shape()
W4 = tf.Variable(tf.truncated_normal([7 * 7 * M, N], stddev=0.1))
print "W4: " , W4.get_shape()
B4 = tf.Variable(tf.ones([N])/10)
print "B4: " , B4.get_shape()
W5 = tf.Variable(tf.truncated_normal([N, 2], stddev=0.1))
print "W5: " , W5.get_shape()
B5 = tf.Variable(tf.ones([2])/10)
print "B5: " , B5.get_shape()
# The model
stride = 1 # output is 28x28
Y1 = tf.nn.relu(tf.nn.conv2d(X, W1, strides=[1, stride, stride, 1], padding='SAME') + B1)
print "Y1: " , Y1.get_shape()
stride = 2 # output is 14x14
Y2 = tf.nn.relu(tf.nn.conv2d(Y1, W2, strides=[1, stride, stride, 1], padding='SAME') + B2)
print "Y2: " , Y2.get_shape()
stride = 2 # output is 7x7
Y3 = tf.nn.relu(tf.nn.conv2d(Y2, W3, strides=[1, stride, stride, 1], padding='SAME') + B3)
print "Y3: " , Y3.get_shape()
# reshape the output from the third convolution for the fully connected layer
#YY = tf.reshape(Y3, shape=[-1, 7 * 7 * M])
YY = tf.reshape(Y3, shape=[-1, 7 * 7 * M])
print "YY: " , YY.get_shape()
Y4 = tf.nn.relu(tf.matmul(YY, W4) + B4)
print "Y4: " , Y4.get_shape()
Ylogits = tf.matmul(Y4, W5) + B5
print "Ylogits: " , Ylogits.get_shape()
Y = tf.nn.softmax(Ylogits)
# cross-entropy loss function (= -sum(Y_i * log(Yi)) ), normalised for batches of 10 images
# TensorFlow provides the softmax_cross_entropy_with_logits function to avoid numerical stability
# problems with log(0) which is NaN
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=Ylogits, labels=Y_)
cross_entropy = tf.reduce_mean(cross_entropy) * 10
# accuracy of the trained model, between 0 (worst) and 1 (best)
correct_prediction = tf.equal(tf.argmax(Y, 1), tf.argmax(Y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
allweights = tf.concat(0, [tf.reshape(W1, [-1]), tf.reshape(W2, [-1]), tf.reshape(W3, [-1]), tf.reshape(W4, [-1]), tf.reshape(W5, [-1])])
allbiases = tf.concat(0, [tf.reshape(B1, [-1]), tf.reshape(B2, [-1]), tf.reshape(B3, [-1]), tf.reshape(B4, [-1]), tf.reshape(B5, [-1])])
# training step, the learning rate is a placeholder
train_step = tf.train.AdamOptimizer(lr).minimize(cross_entropy)
train_images = tf.train.batch([train_image_file], batch_size=train_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * train_batch_size, allow_smaller_final_batch=True)
test_images = tf.train.batch([test_image_file], batch_size=test_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * test_batch_size, allow_smaller_final_batch=True)
train_labels = tf.train.batch([train_image_labels], batch_size=train_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * train_batch_size, enqueue_many=True, allow_smaller_final_batch=True)
test_labels = tf.train.batch([test_image_labels], batch_size=test_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * test_batch_size, enqueue_many=True, allow_smaller_final_batch=True)
init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)
def training_step(i, update_test_data, update_train_data):
train_images_batch = train_images.eval(session=sess)
train_images_batch = np.expand_dims(train_images_batch, axis=(3))
train_labels_batch = train_labels.eval(session=sess)
test_images_batch = test_images.eval(session=sess)
test_images_batch = np.expand_dims(test_images_batch, axis=(3))
test_labels_batch = test_labels.eval(session=sess)
# learning rate decay
max_learning_rate = 0.003
min_learning_rate = 0.0001
decay_speed = 2000.0
learning_rate = min_learning_rate + (max_learning_rate - min_learning_rate) * math.exp(-i/decay_speed)
if update_train_data:
a, c, w, b = sess.run([accuracy, cross_entropy, allweights, allbiases], {X: train_images_batch, Y_: train_labels_batch})
print(str(i) + ": accuracy:" + str(a) + " loss: " + str(c) + " (lr:" + str(learning_rate) + ")")
if update_test_data:
a, c = sess.run([accuracy, cross_entropy], {X: test_images_batch, Y_: test_labels_batch})
print(str(i) + ": ********* epoch " + " ********* test accuracy:" + str(a) + " test loss: " + str(c))
# the backpropagation training step
sess.run(train_step, {X: batch_X, Y_: batch_Y, lr: learning_rate})
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(coord=coord, sess=sess)
for i in range(10000+1):
training_step(i, i % 100 == 0, i % 20 == 0)
coord.request_stop()
coord.join(threads)
结果：
('Dog: ', 12500, ' Cat: ', 12500)
W1: (5, 5, 1, 4)
B1: (4,)
W2: (5, 5, 4, 8)
B2: (8,)
W3: (4, 4, 8, 12)
B3: (12,)
W4: (588, 200)
B4: (200,)
W5: (200, 2)
B5: (2,)
Y1: (?, 224, 224, 4)
Y2: (?, 112, 112, 8)
Y3: (?, 56, 56, 12)
YY: (?, 588)
Y4: (?, 200)
Ylogits: (?, 2)
Traceback (most recent call last):
File "convolutional.py", line 306, in <module>
training_step(i, i % 100 == 0, i % 20 == 0)
File "convolutional.py", line 288, in training_step
a, c, w, b = sess.run([accuracy, cross_entropy, allweights, allbiases], {X: train_images_batch, Y_: train_labels_batch})
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 766, in run
run_metadata_ptr)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 964, in _run
feed_dict_string, options, run_metadata)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1014, in _do_run
target_list, options, run_metadata)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1034, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.InvalidArgumentError: logits and labels must be same size: logits_size=[640,2] labels_size=[10,2]
[[Node: SoftmaxCrossEntropyWithLogits = SoftmaxCrossEntropyWithLogits[T=DT_FLOAT, _device="/job:localhost/replica:0/task:0/cpu:0"](Reshape_1, Reshape_2)]]
Caused by op u'SoftmaxCrossEntropyWithLogits', defined at:
File "convolutional.py", line 229, in <module>
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=Ylogits, labels=Y_)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/ops/nn_ops.py", line 1449, in softmax_cross_entropy_with_logits
precise_logits, labels, name=name)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/ops/gen_nn_ops.py", line 2265, in _softmax_cross_entropy_with_logits
features=features, labels=labels, name=name)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 759, in apply_op
op_def=op_def)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2240, in create_op
original_op=self._default_original_op, op_def=op_def)
File "/home/dragon/.local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1128, in __init__
self._traceback = _extract_stack()
InvalidArgumentError (see above for traceback): logits and labels must be same size: logits_size=[640,2] labels_size=[10,2]
[[Node: SoftmaxCrossEntropyWithLogits = SoftmaxCrossEntropyWithLogits[T=DT_FLOAT, _device="/job:localhost/replica:0/task:0/cpu:0"](Reshape_1, Reshape_2)]]
设法修复它：
import os
import tensorflow as tf
from tensorflow.python.framework import ops
from tensorflow.python.framework import dtypes
import numpy as np
import glob
import fnmatch
import matplotlib.pyplot as plt
from PIL import Image
import random
import threading
import math
tf.set_random_seed(0)
def convertToOneHot(vector, num_classes=None):
assert isinstance(vector, np.ndarray)
assert len(vector) > 0
if num_classes is None:
num_classes = np.max(vector)+1
else:
assert num_classes > 0
assert num_classes >= np.max(vector)
result = np.zeros(shape=(len(vector), num_classes))
result[np.arange(len(vector)), vector] = 1
return result.astype(np.float32)
def make_labels(filenames):
n = len(filenames)
#y = np.zeros((n,2), dtype = np.int32)
#y = np.zeros(shape=[n], dtype = np.float32)
label_y = np.zeros((n,2), dtype = np.float32)
counter = 0
dog = 0
cat = 0
for i in range(n):
# If 'dog' string is in file name assign '1'
if fnmatch.fnmatch(filenames[i], '*dog*'):
label_y[i,0] = 1
#label_y[i] = 1
dog += 1
else:
label_y[i,1] = 1
#label_y[i] = 0
cat += 1
print("Dog: " , dog , " Cat: " , cat)
return label_y
def make_test_labels(filenames):
n = len(filenames)
test_label_y = np.zeros([n], dtype = np.int32)
for i in range(n):
test_label_y[i] = random.randrange(0,2)
one_hot = convertToOneHot(test_label_y)
return one_hot
train_path = "./data/train/*.jpg"
test_path = "./data/test1/*.jpg"
#Training Dataset
train_files = tf.gfile.Glob(train_path)
train_image_labels = make_labels(train_files)
train_filename_queue = tf.train.string_input_producer(train_files, shuffle=False)
train_image_reader = tf.WholeFileReader()
train_image_filename, train_image_file = train_image_reader.read(train_filename_queue)
train_image_file = tf.image.decode_jpeg(train_image_file, 1)
train_image_file = tf.image.resize_images(train_image_file, [224, 224])
train_image_file.set_shape((224, 224, 1))
train_image_file = tf.squeeze(train_image_file)
#Test or Eval Dataset
test_files = tf.gfile.Glob(test_path)
test_image_labels = make_test_labels(test_files)
test_filename_queue = tf.train.string_input_producer(test_files, shuffle=False)
test_image_reader = tf.WholeFileReader()
test_image_filename, test_image_file = test_image_reader.read(test_filename_queue)
test_image_file = tf.image.decode_jpeg(test_image_file, 1)
test_image_file = tf.image.resize_images(test_image_file, [224, 224])
test_image_file.set_shape((224, 224, 1))
test_image_file = tf.squeeze(test_image_file)
train_batch_size = 10
test_batch_size = 2
num_preprocess_threads = 1
min_queue_examples = 256
X = tf.placeholder(tf.float32, [None, 224, 224, 1])
Y_ = tf.placeholder(tf.float32, [None, 2])
lr = tf.placeholder(tf.float32)
pkeep = tf.placeholder(tf.float32)
# three convolutional layers with their channel counts, and a
# fully connected layer (tha last layer has 2 softmax neurons)
K = 4 # first convolutional layer output depth
L = 8 # second convolutional layer output depth
M = 12 # third convolutional layer
N = 200 # fully connected layer
W1 = tf.Variable(tf.truncated_normal([5, 5, 1, K], stddev=0.1)) # 5x5 patch, 1 input channel, K output channels
print "W1: " , W1.get_shape()
B1 = tf.Variable(tf.ones([K])/10)
print "B1: " , B1.get_shape()
W2 = tf.Variable(tf.truncated_normal([5, 5, K, L], stddev=0.1))
print "W2: " , W2.get_shape()
B2 = tf.Variable(tf.ones([L])/10)
print "B2: " , B2.get_shape()
W3 = tf.Variable(tf.truncated_normal([4, 4, L, M], stddev=0.1))
print "W3: " , W3.get_shape()
B3 = tf.Variable(tf.ones([M])/10)
print "B3: " , B3.get_shape()
W4 = tf.Variable(tf.truncated_normal([7 * 7 * M, N], stddev=0.1))
print "W4: " , W4.get_shape()
B4 = tf.Variable(tf.ones([N])/10)
print "B4: " , B4.get_shape()
W5 = tf.Variable(tf.truncated_normal([N, 2], stddev=0.1))
print "W5: " , W5.get_shape()
B5 = tf.Variable(tf.ones([2])/10)
print "B5: " , B5.get_shape()
# The model
stride = 1
Y1 = tf.nn.relu(tf.nn.conv2d(X, W1, strides=[1, stride, stride, 1], padding='SAME') + B1)
print "Y1: " , Y1.get_shape()
stride = 6 #changed from 2 to 6
Y2 = tf.nn.relu(tf.nn.conv2d(Y1, W2, strides=[1, stride, stride, 1], padding='SAME') + B2)
print "Y2: " , Y2.get_shape()
stride = 6 #changed from 2 to 6
Y3 = tf.nn.relu(tf.nn.conv2d(Y2, W3, strides=[1, stride, stride, 1], padding='SAME') + B3)
print "Y3: " , Y3.get_shape()
# reshape the output from the third convolution for the fully connected layer
#YY = tf.reshape(Y3, shape=[-1, 7 * 7 * M])
YY = tf.reshape(Y3, shape=[-1, 7 * 7 * M])
print "YY: " , YY.get_shape()
Y4 = tf.nn.relu(tf.matmul(YY, W4) + B4)
print "Y4: " , Y4.get_shape()
Ylogits = tf.matmul(Y4, W5) + B5
print "Ylogits: " , Ylogits.get_shape()
Y = tf.nn.softmax(Ylogits)
# cross-entropy loss function (= -sum(Y_i * log(Yi)) ), normalised for batches of 10 images
# TensorFlow provides the softmax_cross_entropy_with_logits function to avoid numerical stability
# problems with log(0) which is NaN
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=Ylogits, labels=Y_)
cross_entropy = tf.reduce_mean(cross_entropy) * 10
# accuracy of the trained model, between 0 (worst) and 1 (best)
correct_prediction = tf.equal(tf.argmax(Y, 1), tf.argmax(Y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
allweights = tf.concat(0, [tf.reshape(W1, [-1]), tf.reshape(W2, [-1]), tf.reshape(W3, [-1]), tf.reshape(W4, [-1]), tf.reshape(W5, [-1])])
allbiases = tf.concat(0, [tf.reshape(B1, [-1]), tf.reshape(B2, [-1]), tf.reshape(B3, [-1]), tf.reshape(B4, [-1]), tf.reshape(B5, [-1])])
# training step, the learning rate is a placeholder
train_step = tf.train.AdamOptimizer(lr).minimize(cross_entropy)
train_images = tf.train.batch([train_image_file], batch_size=train_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * train_batch_size, allow_smaller_final_batch=True)
test_images = tf.train.batch([test_image_file], batch_size=test_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * test_batch_size, allow_smaller_final_batch=True)
train_labels = tf.train.batch([train_image_labels], batch_size=train_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * train_batch_size, enqueue_many=True, allow_smaller_final_batch=True)
test_labels = tf.train.batch([test_image_labels], batch_size=test_batch_size, num_threads=num_preprocess_threads, capacity=min_queue_examples + 3 * test_batch_size, enqueue_many=True, allow_smaller_final_batch=True)
init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)
def training_step(i, update_test_data, update_train_data):
train_images_batch = train_images.eval(session=sess)
train_images_batch = np.expand_dims(train_images_batch, axis=(3))
train_labels_batch = train_labels.eval(session=sess)
test_images_batch = test_images.eval(session=sess)
test_images_batch = np.expand_dims(test_images_batch, axis=(3))
test_labels_batch = test_labels.eval(session=sess)
# learning rate decay
max_learning_rate = 0.003
min_learning_rate = 0.0001
decay_speed = 2000.0
learning_rate = min_learning_rate + (max_learning_rate - min_learning_rate) * math.exp(-i/decay_speed)
if update_train_data:
a, c, w, b = sess.run([accuracy, cross_entropy, allweights, allbiases], {X: train_images_batch, Y_: train_labels_batch})
print(str(i) + ": accuracy:" + str(a) + " loss: " + str(c) + " (lr:" + str(learning_rate) + ")")
if update_test_data:
a, c = sess.run([accuracy, cross_entropy], {X: test_images_batch, Y_: test_labels_batch})
print(str(i) + ": ********* epoch " + " ********* test accuracy:" + str(a) + " test loss: " + str(c))
# the backpropagation training step
sess.run(train_step, {X: batch_X, Y_: batch_Y, lr: learning_rate})
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(coord=coord, sess=sess)
for i in range(10000+1):
training_step(i, i % 100 == 0, i % 20 == 0)
coord.request_stop()
coord.join(threads)
结果：
0: accuracy:0.6 loss: 11.1554 (lr:0.003)
0: ********* epoch ********* test accuracy:0.5 test loss: 13.0221
20: accuracy:0.7 loss: 5.93612 (lr:0.00297114451787)
40: accuracy:0.5 loss: 7.40466 (lr:0.00294257615259)
60: accuracy:0.4 loss: 6.7717 (lr:0.00291429204729)
80: accuracy:0.5 loss: 7.07674 (lr:0.00288628937354)
100: accuracy:0.6 loss: 7.0582 (lr:0.00285856533105)
100: ********* epoch ********* test accuracy:0.5 test loss: 7.8989
120: accuracy:0.3 loss: 6.95063 (lr:0.00283111714739)
140: accuracy:0.5 loss: 6.57362 (lr:0.00280394207773)
160: accuracy:0.4 loss: 7.82804 (lr:0.00277703740452)
180: accuracy:0.6 loss: 6.80486 (lr:0.00275040043729)
200: accuracy:0.6 loss: 6.98334 (lr:0.0027240285123)
200: ********* epoch ********* test accuracy:0.5 test loss: 6.71705
220: accuracy:0.9 loss: 4.97062 (lr:0.00269791899236)
240: accuracy:0.5 loss: 5.81752 (lr:0.00267206926648)
260: accuracy:0.6 loss: 6.53955 (lr:0.00264647674967)
280: accuracy:0.3 loss: 8.13235 (lr:0.00262113888266)
300: accuracy:0.5 loss: 6.55135 (lr:0.00259605313163)
300: ********* epoch ********* test accuracy:0.5 test loss: 6.69505
320: accuracy:0.7 loss: 6.20664 (lr:0.002571216988)
340: accuracy:0.6 loss: 6.23909 (lr:0.00254662796813)
360: accuracy:0.8 loss: 4.65571 (lr:0.00252228361309)
380: accuracy:0.5 loss: 8.66489 (lr:0.00249818148844)
400: accuracy:0.6 loss: 6.5182 (lr:0.00247431918393)
400: ********* epoch ********* test accuracy:0.0 test loss: 7.87964
420: accuracy:0.8 loss: 5.66837 (lr:0.00245069431331)
440: accuracy:0.5 loss: 6.87555 (lr:0.00242730451409)
460: accuracy:0.7 loss: 5.9571 (lr:0.00240414744726)
480: accuracy:0.6 loss: 6.63313 (lr:0.00238122079709)
500: accuracy:0.7 loss: 6.31318 (lr:0.00235852227091)
500: ********* epoch ********* test accuracy:0.0 test loss: 8.10029
520: accuracy:0.7 loss: 7.07258 (lr:0.00233604959883)
540: accuracy:0.5 loss: 7.05975 (lr:0.00231380053358)
560: accuracy:0.5 loss: 7.25364 (lr:0.00229177285022)
580: accuracy:0.6 loss: 6.73549 (lr:0.00226996434598)
600: accuracy:0.8 loss: 4.62922 (lr:0.00224837283998)
600: ********* epoch ********* test accuracy:0.5 test loss: 6.42755
620: accuracy:0.6 loss: 6.97526 (lr:0.00222699617305)
640: accuracy:0.7 loss: 5.79574 (lr:0.00220583220751)
660: accuracy:0.7 loss: 6.3626 (lr:0.00218487882695)
680: accuracy:0.8 loss: 7.46478 (lr:0.00216413393601)
700: accuracy:0.6 loss: 5.08642 (lr:0.00214359546018)
700: ********* epoch ********* test accuracy:0.5 test loss: 13.1772
720: accuracy:0.5 loss: 7.97948 (lr:0.00212326134561)
740: accuracy:0.5 loss: 7.21764 (lr:0.00210312955885)
760: accuracy:0.7 loss: 6.36856 (lr:0.00208319808672)
780: accuracy:0.7 loss: 5.55039 (lr:0.00206346493604)
800: accuracy:0.6 loss: 7.82716 (lr:0.0020439281335)
800: ********* epoch ********* test accuracy:0.0 test loss: 8.27366
820: accuracy:0.5 loss: 6.67392 (lr:0.0020245857254)
840: accuracy:0.6 loss: 6.52732 (lr:0.00200543577746)
860: accuracy:0.6 loss: 6.47456 (lr:0.0019864763747)
880: accuracy:0.8 loss: 5.34777 (lr:0.00196770562114)
900: accuracy:0.5 loss: 6.67296 (lr:0.0019491216397)
900: ********* epoch ********* test accuracy:0.0 test loss: 9.64283
920: accuracy:0.7 loss: 6.66137 (lr:0.00193072257197)
940: accuracy:0.9 loss: 5.7024 (lr:0.00191250657802)
960: accuracy:0.7 loss: 6.43189 (lr:0.00189447183624)
980: accuracy:0.7 loss: 5.07373 (lr:0.00187661654313)
1000: accuracy:0.7 loss: 6.38296 (lr:0.00185893891317)
1000: ********* epoch ********* test accuracy:0.5 test loss: 8.52038
1020: accuracy:0.8 loss: 9.55432 (lr:0.00184143717856)
1040: accuracy:0.6 loss: 7.09334 (lr:0.00182410958911)
1060: accuracy:0.4 loss: 7.11957 (lr:0.00180695441207)
1080: accuracy:0.6 loss: 7.0127 (lr:0.00178996993188)
1100: accuracy:0.4 loss: 7.69579 (lr:0.0017731544501)
1100: ********* epoch ********* test accuracy:0.5 test loss: 11.4886
1120: accuracy:0.7 loss: 6.882 (lr:0.00175650628516)
1140: accuracy:0.8 loss: 5.1929 (lr:0.00174002377223)
1160: accuracy:0.3 loss: 6.76535 (lr:0.00172370526304)
1180: accuracy:0.5 loss: 7.55367 (lr:0.00170754912573)
1200: accuracy:0.7 loss: 5.56166 (lr:0.00169155374467)
1200: ********* epoch ********* test accuracy:0.5 test loss: 7.25217
1220: accuracy:0.8 loss: 5.60191 (lr:0.00167571752032)
1240: accuracy:0.6 loss: 6.82252 (lr:0.00166003886902)
1260: accuracy:0.9 loss: 5.644 (lr:0.00164451622292)
1280: accuracy:0.8 loss: 5.7729 (lr:0.00162914802972)
1300: accuracy:0.7 loss: 5.87659 (lr:0.00161393275261)
1300: ********* epoch ********* test accuracy:0.5 test loss: 5.97413
1320: accuracy:0.5 loss: 7.00393 (lr:0.00159886887003)
1340: accuracy:0.5 loss: 6.74683 (lr:0.00158395487558)
1360: accuracy:0.5 loss: 6.34526 (lr:0.00156918927786)
1380: accuracy:0.5 loss: 7.20938 (lr:0.00155457060029)
1400: accuracy:0.5 loss: 6.90222 (lr:0.001540097381)
1400: ********* epoch ********* test accuracy:1.0 test loss: 5.80986
1420: accuracy:0.7 loss: 6.27892 (lr:0.00152576817264)
1440: accuracy:0.6 loss: 7.12203 (lr:0.00151158154228)
1460: accuracy:0.6 loss: 6.3593 (lr:0.00149753607126)
1480: accuracy:0.5 loss: 6.73234 (lr:0.00148363035501)
1500: accuracy:0.7 loss: 5.93877 (lr:0.00146986300295)
1500: ********* epoch ********* test accuracy:0.0 test loss: 13.2649
1520: accuracy:0.5 loss: 6.56755 (lr:0.00145623263833)
1540: accuracy:0.2 loss: 11.2914 (lr:0.0014427378981)
1560: accuracy:0.7 loss: 6.40623 (lr:0.00142937743279)
1580: accuracy:0.8 loss: 5.21643 (lr:0.00141614990632)
1600: accuracy:0.6 loss: 7.13112 (lr:0.00140305399594)
1600: ********* epoch ********* test accuracy:0.5 test loss: 7.00448
1620: accuracy:0.8 loss: 6.04726 (lr:0.00139008839205)
1640: accuracy:0.6 loss: 5.6949 (lr:0.00137725179807)
1660: accuracy:0.4 loss: 8.35133 (lr:0.00136454293033)
1680: accuracy:0.7 loss: 5.67702 (lr:0.00135196051794)
1700: accuracy:0.7 loss: 5.75319 (lr:0.00133950330265)
1700: ********* epoch ********* test accuracy:0.5 test loss: 4.69712
1720: accuracy:0.7 loss: 5.62581 (lr:0.00132717003872)
1740: accuracy:0.5 loss: 6.94164 (lr:0.00131495949282)
1760: accuracy:0.8 loss: 6.81993 (lr:0.00130287044388)
1780: accuracy:0.6 loss: 6.46852 (lr:0.00129090168298)
1800: accuracy:0.7 loss: 5.52436 (lr:0.00127905201325)
1800: ********* epoch ********* test accuracy:1.0 test loss: 5.64173
1820: accuracy:0.6 loss: 5.64919 (lr:0.0012673202497)
1840: accuracy:0.6 loss: 6.43817 (lr:0.00125570521915)
1860: accuracy:0.6 loss: 6.09095 (lr:0.00124420576008)
1880: accuracy:0.7 loss: 5.98002 (lr:0.00123282072254)
1900: accuracy:0.8 loss: 5.04342 (lr:0.00122154896802)
1900: ********* epoch ********* test accuracy:0.5 test loss: 5.80969
1920: accuracy:0.5 loss: 6.73896 (lr:0.00121038936933)
1940: accuracy:0.6 loss: 5.74824 (lr:0.0011993408105)
1960: accuracy:0.9 loss: 5.07425 (lr:0.00118840218667)
1980: accuracy:0.8 loss: 5.07815 (lr:0.00117757240396)
2000: accuracy:0.6 loss: 6.49868 (lr:0.0011668503794)
2000: ********* epoch ********* test accuracy:0.5 test loss: 5.17815
2020: accuracy:0.7 loss: 6.32198 (lr:0.00115623504076)
2040: accuracy:0.5 loss: 6.81228 (lr:0.0011457253265)
2060: accuracy:0.7 loss: 6.27763 (lr:0.00113532018565)
2080: accuracy:0.4 loss: 6.48217 (lr:0.00112501857768)
2100: accuracy:0.8 loss: 5.47317 (lr:0.00111481947242)
2100: ********* epoch ********* test accuracy:0.5 test loss: 6.97906
2120: accuracy:0.7 loss: 6.64062 (lr:0.00110472184996)
2140: accuracy:0.4 loss: 7.58298 (lr:0.00109472470051)
2160: accuracy:0.5 loss: 7.05855 (lr:0.00108482702437)
2180: accuracy:0.7 loss: 5.3101 (lr:0.00107502783175)
2200: accuracy:0.4 loss: 7.99556 (lr:0.00106532614272)
2200: ********* epoch ********* test accuracy:0.0 test loss: 30.0326
2220: accuracy:0.5 loss: 5.50862 (lr:0.00105572098712)
2240: accuracy:0.7 loss: 7.57991 (lr:0.00104621140441)
2260: accuracy:0.9 loss: 4.57821 (lr:0.00103679644362)
2280: accuracy:0.5 loss: 6.55349 (lr:0.00102747516327)
2300: accuracy:0.4 loss: 7.27755 (lr:0.0010182466312)
2300: ********* epoch ********* test accuracy:0.5 test loss: 8.20956
2320: accuracy:0.7 loss: 5.98563 (lr:0.00100910992456)
2340: accuracy:0.7 loss: 5.72076 (lr:0.00100006412967)
2360: accuracy:0.7 loss: 5.22523 (lr:0.000991108341943)
2380: accuracy:0.8 loss: 4.87689 (lr:0.000982241665793)
2400: accuracy:0.8 loss: 5.14703 (lr:0.000973463214545)
2400: ********* epoch ********* test accuracy:0.5 test loss: 6.0281
2420: accuracy:0.7 loss: 6.03927 (lr:0.000964772110347)
2440: accuracy:0.7 loss: 4.91816 (lr:0.00095616748408)
2460: accuracy:0.5 loss: 6.61033 (lr:0.000947648475274)
2480: accuracy:0.7 loss: 4.98853 (lr:0.000939214232023)
2500: accuracy:0.5 loss: 6.02745 (lr:0.000930863910895)
2500: ********* epoch ********* test accuracy:0.5 test loss: 6.64094
2520: accuracy:0.8 loss: 5.77956 (lr:0.000922596676849)
2540: accuracy:0.5 loss: 7.03326 (lr:0.000914411703157)
2560: accuracy:0.9 loss: 3.64503 (lr:0.000906308171314)
2580: accuracy:0.5 loss: 7.78428 (lr:0.00089828527096)
好像准确率不太好…………是不是要加层数？
还是哪里出了问题？
谢谢