def train_cnn(x_all_data, y_all_data):
#output = create_cnn(3)
output = create_cnn_3muti3(5)
print_info("created cnn ...")
print_info("start train ...")
loss = tf.reduce_mean(
tf.nn.sigmoid_cross_entropy_with_logits(labels=Y, logits=output))
optimizer = tf.train.AdadeltaOptimizer(learning_rate=0.001).minimize(loss)
max_idx_p = tf.argmax(output, 1)
max_idx_l = tf.argmax(Y, 1)
correct_pred = tf.equal(max_idx_p, max_idx_l)
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
saver = tf.train.Saver(max_to_keep=3)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
step = 0
while True:
batch_x, batch_y = get_next_batch(x_all_data, y_all_data)
_, loss_ = sess.run([optimizer, loss],
feed_dict={
X: batch_x,
Y: batch_y,
keep_prob: 0.75
})
print(step, loss_)
'''
out = sess.run([output],feed_dict={X:batch_x,Y:batch_y,keep_prob:0.75})
print("out:",out)
y = sess.run([Y],{X:batch_x,Y:batch_y,keep_prob:0.75})
print("y:",y)
max_p = sess.run([max_idx_p],feed_dict={X:batch_x,Y:batch_y,keep_prob:0.75})
max_l = sess.run([max_idx_l],feed_dict={X:batch_x,Y:batch_y,keep_prob:0.75})
print("max_p:",max_p)
print("max_l",max_l)
break
'''
if step % 10 == 0:
batch_x_test, batch_y_test = get_next_batch(
x_all_data, y_all_data)
acc = sess.run(accuracy,
feed_dict={
X: batch_x_test,
Y: batch_y_test,
keep_prob: 1.0
})
print(step, acc)
#break
if acc > 0.9:
saver.save(sess, "./model_3/cnn.model", global_step=step)
if acc > 0.98:
saver.save(sess, "./model_3/cnn.model", global_step=step)
break
step += 1
def read_test_data(file_in, size):
print_info("start read test data...")
data = read_signal_txt(file_in, 400, 400, 224)
print_info("end read test data...")
test_data = []
x, y, z = data.shape
for j in range(size, y + 1):
for i in range(size, x + 1):
test_cur = data[i - size:i, j - size:j, :]
test_data.append(test_cur)
#test_data = np.array(test_data)
return test_data
def read_test_data(file_in,size):
print_info("start read test data...")
data = read_signal_txt(file_in,2100,510,224)
print_info("end read test data...")
test_data = []
x,y,z = data.shape
for j in range(size,y+1):
for i in range(size,x+1):
test_cur = data[i-size:i,j-size:j,:] #以i为中心的一个矩形区域数据，相当于存储为一个batch，
# 在CNN中对每个batch进行test
test_data.append(test_cur)
#test_data = np.array(test_data)
return test_data
def read_all_train_data():
x_data = []
y_data = []
data_dir = ('E:/Imaging/CNNROI/5_%d_%d.txt')
for i in range(5):
for j in range(1, 21):
cur_txt = data_dir % (i, j)
cur_data = read_signal_txt(cur_txt, 5, 5, 224)
if i == 0:
t = [1, 0, 0, 0, 0]
if i == 1:
t = [0, 1, 0, 0, 0]
if i == 2:
t = [0, 0, 1, 0, 0]
if i == 3:
t = [0, 0, 0, 1, 0]
if i == 4:
t = [0, 0, 0, 0, 1]
y_data.append(t)
x_data.append(cur_data)
print_info("read done!")
x_data = np.array(x_data, dtype=np.float32)
y_data = np.array(y_data, dtype=np.float32)
return x_data, y_data
def read_all_train_data():
x_data = []
y_data = []
data_dir = ('F:/Python/workshop/fishc/CNNROI/5_%d_%d.txt')
for i in range(5):
for j in range(1,21):
cur_txt = data_dir%(i,j)
cur_data = read_signal_txt(cur_txt,5,5,224)
if i == 0: #以文件名中的编号作为区分类别的label，即训练中的Y值
t = [1,0,0,0,0]
if i == 1:
t = [0,1,0,0,0]
if i == 2:
t = [0,0,1,0,0]
if i == 3:
t = [0,0,0,1,0]
if i == 4:
t = [0,0,0,0,1]
y_data.append(t)
x_data.append(cur_data)
print_info("read done!")
x_data = np.array(x_data,dtype = np.float32)
y_data = np.array(y_data,dtype = np.float32)
return x_data,y_data