No description, website, or topics provided.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Semantic Segmentation in Advanced Deep Learning

In this project, I've built a Fully Convolutional Network in order to segment camera images, whether an area is a road or not. The camera is mounted inside a vehicle. The approach of the semantic segmentation is based on the this publication.

After running this project, the main program will generate a logs directory for the visualization using tensorboard and a runs directory which contains the segmented images from the input datasets.

Network Architecture

Tensorflow provides a tool tensorboard to visualize the Convolutional Network. The below figure shows the overview of the VGG16 architecture and the additional upsampling and skipping layers in this project:



Based on my limited experiments, I found out that the number of the epochs returns more visual difference on the images. Applying these fix hyperparameters:

  • learning rate: 0.001
  • keep_prob: 0.5

the bellow tables shows the difference of the result on some images.

Epoch 10 Epoch 40
Epoch 10 Epoch 40
Epoch 10 Epoch 40
Epoch 10 Epoch 40
Epoch 10 Epoch 40

As shown in the above table, the higher the number of epoch, the better is the segmentation result. The cross entropy loss reaches 0.02553. To support this observation I add the cross entropy loss to the tensorflow summary. The graph of the cross entropy loss can be visualized as this below figure:

loss graph

Udacity's Project Introduction

In this project, you'll label the pixels of a road in images using a Fully Convolutional Network (FCN).


Frameworks and Packages

Make sure you have the following is installed:


Download the Kitti Road dataset from here. Extract the dataset in the data folder. This will create the folder data_road with all the training a test images.



Implement the code in the module indicated by the "TODO" comments. The comments indicated with "OPTIONAL" tag are not required to complete.


Run the following command to run the project:


Note If running this in Jupyter Notebook system messages, such as those regarding test status, may appear in the terminal rather than the notebook.


  1. Ensure you've passed all the unit tests.
  2. Ensure you pass all points on the rubric.
  3. Submit the following in a zip file.
  • Newest inference images from runs folder (all images from the most recent run)

How to write a README

A well written README file can enhance your project and portfolio. Develop your abilities to create professional README files by completing this free course.