matplotlib example tensorboard (snippet)

Python matplotlib example 'tensorboard'

Modules used in program:

• import os
• import numpy as np
• import matplotlib.pyplot as plt
• import tensorflow as tf

python tensorboard

Python matplotlib example: tensorboard

import tensorflow as tf
import matplotlib.pyplot as plt
import numpy as np
import os
%matplotlib inline

tf.reset_default_graph()

image_dir = "./image_png"
image_names = os.listdir(image_dir)
images = [os.path.join(image_dir, name) for name in image_names]

imagename_queue = tf.train.string_input_producer(images, shuffle=False)
labelname_queue = tf.train.string_input_producer(['./label_csv/Label.csv'], shuffle=False)


image_reader = tf.WholeFileReader()
key, raw_img = image_reader.read(imagename_queue)

csv_reader = tf.TextLineReader()
key, raw_txt = csv_reader.read(labelname_queue)

png_image = tf.image.decode_png(raw_img)
csv_label = tf.decode_csv(raw_txt, record_defaults=[[0]])

png_image = tf.reduce_mean(png_image, axis=2)
png_image = tf.reshape(png_image, [61, 49, 1])

#x, y_ = tf.train.batch([png_image, csv_label], 32)
x, y_ = tf.train.shuffle_batch([png_image, csv_label], batch_size=32, capacity=50000, min_after_dequeue=10000)
x = tf.cast(x, tf.float32)

y_ = tf.reshape(y_, [-1])
y_ = tf.one_hot(y_,depth=3, on_value=1.0, off_value=0.0, axis=-1, dtype=tf.float32)
#y_ = tf.reduce_mean(y_,axis=2)



conv1 = tf.layers.conv2d(x, filters=13, kernel_size=[3,3], activation=tf.nn.relu, padding="SAME")

conv2 = tf.layers.conv2d(conv1, filters=36, kernel_size=[3,3], activation=tf.nn.relu, padding="SAME")
tf.summary.image('feature_map_conv2', tf.reduce_mean(conv2, axis=3, keep_dims=True))

pool2 = tf.layers.max_pooling2d(conv2, pool_size=[2,2], strides=[2,2])
#conv image check
tf.summary.image('feature_map_pool2', tf.reduce_mean(pool2, axis=3, keep_dims=True))


conv3 = tf.layers.conv2d(pool2, filters=50, kernel_size=[3,3], activation=tf.nn.relu, padding="SAME")
pool3 = tf.layers.max_pooling2d(conv3, pool_size=[2,2], strides=[2,2])

conv4 = tf.layers.conv2d(pool3, filters=20, kernel_size=[3,3], activation=tf.nn.relu, padding="SAME")
pool4 = tf.layers.max_pooling2d(conv4, pool_size=[2,2], strides=[2,2])

fc_input_size = int(pool4.shape[1])*int(pool4.shape[2])*int(pool4.shape[3])
flat = tf.reshape(pool4, shape=[-1, fc_input_size])

drop_prob = tf.placeholder(tf.float32)

fc1 = tf.layers.dense(flat, units=300)
dropout1 = tf.layers.dropout(fc1, drop_prob)

fc2 = tf.layers.dense(dropout1, units=100)
dropout2 = tf.layers.dropout(fc2, drop_prob)

out = tf.layers.dense(dropout2, units=3)

pred = tf.nn.softmax(out)
accuracy = tf.metrics.accuracy(tf.argmax(y_,1), tf.argmax(pred,1))

for i, var in enumerate(tf.trainable_variables()):
    tf.summary.histogram(var.name,var)
#    tf.summary.histogram("variable_{}".format(i), var)

loss = tf.losses.softmax_cross_entropy(y_, out)
tf.summary.scalar("loss",loss)

train_op = tf.train.GradientDescentOptimizer(1e-6).minimize(loss)

merged = tf.summary.merge_all()

with tf.Session() as sess:
    coord = tf.train.Coordinator()
    thread = tf.train.start_queue_runners(sess, coord)    
    writer = tf.summary.FileWriter('./logs/', sess.graph)
    sess.run(tf.global_variables_initializer())
    sess.run(tf.local_variables_initializer())

    for i in range(100):    
        _, _loss, _summaries = sess.run([train_op, loss, merged], feed_dict={drop_prob:0.7})
        writer.add_summary(_summaries, i)
        if(i%10 == 0):
            _pred, _acc = sess.run([pred, accuracy], feed_dict={drop_prob:1.0})
            print('loss: {}, accuracy: {}'.format(_loss, _acc))
            print('predicton1:', _pred[0])
            print('predicton2:', _pred[1])
            print('predicton3:', _pred[2])


    coord.request_stop()
    coord.join(thread)

Python links

• Learn Python: https://pythonbasics.org/
• Python Tutorial: https://pythonprogramminglanguage.com