FCN - Full Code

main.py

#--------------------------
# USER-SPECIFIED DATA
#--------------------------

# Tune these parameters

num_classes = 2
image_shape = (160, 576)
EPOCHS = 40
BATCH_SIZE = 16
DROPOUT = 0.75

# Specify these directory paths

data_dir = './data'
runs_dir = './runs'
training_dir ='./data/data_road/training'
vgg_path = './data/vgg'

#--------------------------
# PLACEHOLDER TENSORS
#--------------------------

correct_label = tf.placeholder(tf.float32, [None, IMAGE_SHAPE[0], IMAGE_SHAPE[1], NUMBER_OF_CLASSES])
learning_rate = tf.placeholder(tf.float32)
keep_prob = tf.placeholder(tf.float32)

#--------------------------
# FUNCTIONS
#--------------------------

def load_vgg(sess, vgg_path):
  
  # load the model and weights
  model = tf.saved_model.loader.load(sess, ['vgg16'], vgg_path)

  # Get Tensors to be returned from graph
  graph = tf.get_default_graph()
  image_input = graph.get_tensor_by_name('image_input:0')
  keep_prob = graph.get_tensor_by_name('keep_prob:0')
  layer3 = graph.get_tensor_by_name('layer3_out:0')
  layer4 = graph.get_tensor_by_name('layer4_out:0')
  layer7 = graph.get_tensor_by_name('layer7_out:0')

  return image_input, keep_prob, layer3, layer4, layer7

def layers(vgg_layer3_out, vgg_layer4_out, vgg_layer7_out, num_classes):
   
    # Use a shorter variable name for simplicity
    layer3, layer4, layer7 = vgg_layer3_out, vgg_layer4_out, vgg_layer7_out

    # Apply 1x1 convolution in place of fully connected layer
    fcn8 = tf.layers.conv2d(layer7, filters=num_classes, kernel_size=1, name="fcn8")

    # Upsample fcn8 with size depth=(4096?) to match size of layer 4 so that we can add skip connection with 4th layer
    fcn9 = tf.layers.conv2d_transpose(fcn8, filters=layer4.get_shape().as_list()[-1],
    kernel_size=4, strides=(2, 2), padding='SAME', name="fcn9")

    # Add a skip connection between current final layer fcn8 and 4th layer
    fcn9_skip_connected = tf.add(fcn9, layer4, name="fcn9_plus_vgg_layer4")

    # Upsample again
    fcn10 = tf.layers.conv2d_transpose(fcn9_skip_connected, filters=layer3.get_shape().as_list()[-1],
    kernel_size=4, strides=(2, 2), padding='SAME', name="fcn10_conv2d")

    # Add skip connection
    fcn10_skip_connected = tf.add(fcn10, layer3, name="fcn10_plus_vgg_layer3")

    # Upsample again
    fcn11 = tf.layers.conv2d_transpose(fcn10_skip_connected, filters=num_classes,
    kernel_size=16, strides=(8, 8), padding='SAME', name="fcn11")

    return fcn11

def optimize(nn_last_layer, correct_label, learning_rate, num_classes):
  
  # Reshape 4D tensors to 2D, each row represents a pixel, each column a class
  logits = tf.reshape(nn_last_layer, (-1, num_classes), name="fcn_logits")
  correct_label_reshaped = tf.reshape(correct_label, (-1, num_classes))

  # Calculate distance from actual labels using cross entropy
  cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=correct_label_reshaped[:])
  # Take mean for total loss
  loss_op = tf.reduce_mean(cross_entropy, name="fcn_loss")

  # The model implements this operation to find the weights/parameters that would yield correct pixel labels
  train_op = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(loss_op, name="fcn_train_op")

  return logits, train_op, loss_op

def train_nn(sess, epochs, batch_size, get_batches_fn, train_op,
             cross_entropy_loss, input_image,
             correct_label, keep_prob, learning_rate):

  keep_prob_value = 0.5
  learning_rate_value = 0.001
  for epoch in range(epochs):
      # Create function to get batches
      total_loss = 0
      for X_batch, gt_batch in get_batches_fn(batch_size):

          loss, _ = sess.run([cross_entropy_loss, train_op],
          feed_dict={input_image: X_batch, correct_label: gt_batch,
          keep_prob: keep_prob_value, learning_rate:learning_rate_value})

          total_loss += loss;

      print("EPOCH {} ...".format(epoch + 1))
      print("Loss = {:.3f}".format(total_loss))
      print()

def run():
  
  # Download pretrained vgg model
  helper.maybe_download_pretrained_vgg(data_dir)

  # A function to get batches
  get_batches_fn = helper.gen_batch_function(training_dir, image_shape)
  
  with tf.Session() as session:
        
    # Returns the three layers, keep probability and input layer from the vgg architecture
    image_input, keep_prob, layer3, layer4, layer7 = load_vgg(session, vgg_path)

    # The resulting network architecture from adding a decoder on top of the given vgg model
    model_output = layers(layer3, layer4, layer7, num_classes)

    # Returns the output logits, training operation and cost operation to be used
    # - logits: each row represents a pixel, each column a class
    # - train_op: function used to get the right parameters to the model to correctly label the pixels
    # - cross_entropy_loss: function outputting the cost which we are minimizing, lower cost should yield higher accuracy
    logits, train_op, cross_entropy_loss = optimize(model_output, correct_label, learning_rate, num_classes)
    
    # Initialize all variables
    session.run(tf.global_variables_initializer())
    session.run(tf.local_variables_initializer())

    print("Model build successful, starting training")

    # Train the neural network
    train_nn(session, EPOCHS, BATCH_SIZE, get_batches_fn, 
             train_op, cross_entropy_loss, image_input,
             correct_label, keep_prob, learning_rate)

    # Run the model with the test images and save each painted output image (roads painted green)
    helper.save_inference_samples(runs_dir, data_dir, session, image_shape, logits, keep_prob, image_input)
    
    print("All done!")

#--------------------------
# MAIN
#--------------------------
if __name__ == '__main__':
    run()

could someone pass the url of the "helper" library that was used in this code?

I know it's been a while since you asked this question, but I found the answer, and though I would share it here to help someone else.

He is solving the problem offered in this repo:
https://github.com/udacity/CarND-Semantic-Segmentation

another solution is here:
https://gist.github.com/lianyi/a5ba8d84f5b68401c2313b05e020b062

Great tutorial, anyaways we are facing a problem while opening a picture. It doesn’t happend for every picture but always for the same picture.

In get_batches_fn()

image = scipy.misc.imresize(scipy.misc.imread(image_file), image_shape)

OSError: cannot identify image file ‘./drive/My Drive/…./file.png

Do you have any solution for this ?
thank you

Hi, Thanks for this Great tutorial, I'm trying to run the file main.py, and it give me the error :
Traceback (most recent call last):

File "main.py", line 156, in
run()
File "main.py", line 117, in run
helper.maybe_download_pretrained_vgg(data_dir)
NameError: name 'helper' is not defined

I need help please!!!

Please how do i can run this code! Can someone share the helper file.

After training, my "run" file is empty, and I noticed that the loss in every epoch is 0. Can someone tell me why that might be

correct_label = tf.placeholder(tf.float32, [None, IMAGE_SHAPE[0], IMAGE_SHAPE[1], NUMBER_OF_CLASSES])
I guess IMAGE_SHAPE is a typo and same as image_shape ?

In 'load_vgg' a 'model' variable is defined but never used. why ?

Hello
I tried executing your code but i'm getting allocation of X.. exceeded 10% of memeory error
please help

@Amithsai This error is showing because your training data is not fitting to the memory. You can resolve this error by reducing the batch size you are giving to the network while training the network. Even though, you are facing this issue does your network still runs?

hi there, why must the image shape fixed as (160,576) i thought that the FCN can accept variable sized input?

helper.py

Are vgg parameters also trained?

@chunchet-ng Did you solve your question？

Is there anyone who faced with that problem : I train the model, there isnt any problem, the loss starts from nearly 68.000, then decrease directly to 0 at the 2. epoch. I train the model for 40 epoch and at the test phase, I don't see any segmentation result. The output is just the same with test images without any segmentation prediction.