main.py

# !pip install tqdm
# !pip install moviepy

# import imageio
# imageio.plugins.ffmpeg.download()


import os.path
import tensorflow as tf
import helper
import warnings
from distutils.version import LooseVersion
import project_tests as tests
import numpy as np

import matplotlib.pyplot as plt
import scipy
get_ipython().magic('matplotlib inline')
from moviepy.editor import VideoFileClip



# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer.  You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))

# Check for a GPU
if not tf.test.gpu_device_name():
    warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))


def load_vgg(sess, vgg_path):
    """
    Load Pretrained VGG Model into TensorFlow.
    :param sess: TensorFlow Session
    :param vgg_path: Path to vgg folder, containing "variables/" and "saved_model.pb"
    :return: Tuple of Tensors from VGG model (image_input, keep_prob, layer3_out, layer4_out, layer7_out)
    """
    # Implement function
    #   Use tf.saved_model.loader.load to load the model and weights
    vgg_tag = 'vgg16'
    vgg_input_tensor_name = 'image_input:0'
    vgg_keep_prob_tensor_name = 'keep_prob:0'
    vgg_layer3_out_tensor_name = 'layer3_out:0'
    vgg_layer4_out_tensor_name = 'layer4_out:0'
    vgg_layer7_out_tensor_name = 'layer7_out:0'
    
    tf.saved_model.loader.load(sess, ['vgg16'], vgg_path)
    graph = tf.get_default_graph()
    image_input = graph.get_tensor_by_name(vgg_input_tensor_name)
    keep_prob = graph.get_tensor_by_name(vgg_keep_prob_tensor_name)
    layer3 = graph.get_tensor_by_name(vgg_layer3_out_tensor_name)
    layer4 = graph.get_tensor_by_name(vgg_layer4_out_tensor_name)
    layer7 = graph.get_tensor_by_name(vgg_layer7_out_tensor_name)
    
    return image_input, keep_prob, layer3, layer4, layer7
tests.test_load_vgg(load_vgg, tf)


def layers(vgg_layer3_out, vgg_layer4_out, vgg_layer7_out, num_classes):
    """
    Create the layers for a fully convolutional network.  Build skip-layers using the vgg layers.
    :param vgg_layer7_out: TF Tensor for VGG Layer 3 output
    :param vgg_layer4_out: TF Tensor for VGG Layer 4 output
    :param vgg_layer3_out: TF Tensor for VGG Layer 7 output
    :param num_classes: Number of classes to classify
    :return: The Tensor for the last layer of output
    """
    
    kernel_regularizer=tf.contrib.layers.l2_regularizer(1e-3)
    kernel_initializer=tf.truncated_normal_initializer(stddev=0.01)
    
    
    # TODO: Implement function
    conv1x1= tf.layers.conv2d(vgg_layer7_out, filters = num_classes,kernel_size=1, strides = (1,1), padding = 'same',
                             kernel_regularizer=kernel_regularizer,kernel_initializer = kernel_initializer)
    up_sampled_1 = tf.layers.conv2d_transpose(conv1x1, num_classes,4, 2, padding='same',
                              kernel_regularizer=kernel_regularizer,kernel_initializer = kernel_initializer)
    #print(conv1x1.get_shape,up_sampled_1.get_shape)
    #adding the skipped connections
    
    vgg_layer4_1x1 = tf.layers.conv2d(vgg_layer4_out, num_classes,1, strides=(1,1), padding= 'same',
                             kernel_regularizer=kernel_regularizer,kernel_initializer = kernel_initializer)

    skip_1 = tf.add(up_sampled_1, vgg_layer4_1x1)
    
    up_sampled_2 = tf.layers.conv2d_transpose(skip_1, num_classes,4,2, padding='same',
                             kernel_regularizer=kernel_regularizer,kernel_initializer = kernel_initializer)
    #print (up_sampled_2.get_shape)
    
    vgg_layer3_1x1 = tf.layers.conv2d(vgg_layer3_out, num_classes,1, strides=(1,1), padding= 'same',
                             kernel_regularizer=kernel_regularizer,kernel_initializer = kernel_initializer)
    skip_2 = tf.add(up_sampled_2, vgg_layer3_1x1)
    
    up_sampled_3 = tf.layers.conv2d_transpose(skip_2, num_classes, 16,8, padding= 'same',
                                             kernel_regularizer=kernel_regularizer,kernel_initializer = kernel_initializer)
    
    return up_sampled_3
tests.test_layers(layers)


def optimize(nn_last_layer, correct_label, learning_rate, num_classes):
    """
    Build the TensorFLow loss and optimizer operations.
    :param nn_last_layer: TF Tensor of the last layer in the neural network
    :param correct_label: TF Placeholder for the correct label image
    :param learning_rate: TF Placeholder for the learning rate
    :param num_classes: Number of classes to classify
    :return: Tuple of (logits, train_op, cross_entropy_loss)
    """
    # Implement function
    logits = tf.reshape(nn_last_layer, (-1, num_classes))
    labels_reshaped = tf.reshape(correct_label, (-1, num_classes))
    cross_entropy_loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels= labels_reshaped))
    
    optimizer = tf.train.AdamOptimizer(learning_rate = learning_rate)
    
    training_operation = optimizer.minimize(cross_entropy_loss)
    
    return logits, training_operation, cross_entropy_loss
print(optimize)
tests.test_optimize(optimize)


def train_nn(sess, epochs, batch_size, get_batches_fn, train_op, cross_entropy_loss, input_image,
             correct_label, keep_prob, learning_rate):
    """
    Train neural network and print out the loss during training.
    :param sess: TF Session
    :param epochs: Number of epochs
    :param batch_size: Batch size
    :param get_batches_fn: Function to get batches of training data.  Call using get_batches_fn(batch_size)
    :param train_op: TF Operation to train the neural network
    :param cross_entropy_loss: TF Tensor for the amount of loss
    :param input_image: TF Placeholder for input images
    :param correct_label: TF Placeholder for label images
    :param keep_prob: TF Placeholder for dropout keep probability
    :param learning_rate: TF Placeholder for learning rate
    """
    l_r = 0.00005
    kp_pr = 0.75
    import numpy as np
    sess.run(tf.global_variables_initializer())
    for i in range(epochs):
        for images, labels in get_batches_fn(batch_size):
            _, loss = sess.run([train_op, cross_entropy_loss], feed_dict ={input_image:images, correct_label:labels, learning_rate:l_r, keep_prob:kp_pr})
             
            print("cost =", "{:.9f}".format(loss))
        print("Epoch:", '%04d | ' % (i+1), "cost =", "{:.9f}".format(loss))

tests.test_train_nn(train_nn)


def run():
    num_classes = 2
    image_shape = (160, 576)
    epochs = 15
    learning_rate = tf.placeholder(tf.float32)

    batch_size = 8
    correct_label = tf.placeholder(tf.int64, (None,None,None, 2))
    data_dir = './data'
    runs_dir = './runs'
    tests.test_for_kitti_dataset(data_dir)

    # Download pretrained vgg model
    helper.maybe_download_pretrained_vgg(data_dir)

    # OPTIONAL: Train and Inference on the cityscapes dataset instead of the Kitti dataset.
    # You'll need a GPU with at least 10 teraFLOPS to train on.
    #  https://www.cityscapes-dataset.com/

    with tf.Session() as sess:
        # Path to vgg model
        vgg_path = os.path.join(data_dir, 'vgg')
        # Create function to get batches
        get_batches_fn = helper.gen_batch_function(os.path.join(data_dir, 'data_road/training'), image_shape)

        # OPTIONAL: Augment Images for better results
        #  https://datascience.stackexchange.com/questions/5224/how-to-prepare-augment-images-for-neural-network

        # Build NN using load_vgg, layers, and optimize function
        image_input, keep_prob, layer3, layer4, layer7 = load_vgg(sess, vgg_path)
        nn_last_layer = layers(layer3, layer4, layer7, num_classes)
        logits, train_op, cross_entropy_loss = optimize(nn_last_layer, correct_label, learning_rate, num_classes)

        #  Train NN using the train_nn function
        train_nn(sess, epochs, batch_size, get_batches_fn, train_op, cross_entropy_loss, image_input,
             correct_label, keep_prob, learning_rate)

        # Save inference data using helper.save_inference_samples
        helper.save_inference_samples(runs_dir, data_dir, sess, image_shape, logits, keep_prob, image_input)


        #  Apply the trained model to a video
        clip1 = VideoFileClip("Clip_KITTI_dataset.mp4")
        lambda_for_func = lambda imgg: helper.single_img_inference(imgg, image_shape, sess, image_input, logits, keep_prob)
        out_clip = clip1.fl_image(lambda_for_func)
        
        output = 'Processed_Clip_KITTI_dataset.mp4'
        out_clip.write_videofile(output, audio=False)

if __name__ == '__main__':
    run()