In this project a Fully Convolutional Network (FCN) is trained to label the pixels of a road in images.
Make sure you have the following is installed:
The kitti dataset is used to train the model. First try to download Kitti Road dataset with this command:
wget http://kitti.is.tue.mpg.de/kitti/data_road.zip
If that link has expired request download link to Kitti Road dataset from here.
Once downloaded extract the dataset in the data
folder. This will create the folder data_road
with all the training a test images.
Use the following command to run the project:
python3 main.py
Note If running this in Jupyter Notebook system messages, such as those regarding test status, may appear in the terminal rather than the notebook.
Note If running on AWS here are some helpful command templates:
- You need to change the IP address in the link to your specific instance which can be found on AWS EC2 management console Instances group by clicking "Connect" on the instance:
- You need the AWS key file, e.g. "demo.pem", in your current directory
- SSH connection to AWS instance:
ssh -i "demo.pem" [email protected]
- Copy data_road.zip from local machine to instance:
scp -i "demo.pem" '/mnt/c/Users/nxa09564/OneDrive - NXP/Udacity/SDCND/projects/term3/data_road.zip' [email protected]:~/CarND-Semantic-Segmentation/data
- Copy vgg.zip from local machine to instance:
scp -i "demo.pem" '/mnt/c/Users/nxa09564/OneDrive - NXP/Udacity/SDCND/projects/term3/vgg.zip' [email protected]:~/CarND-Semantic-Segmentation/data
- Copy run output images from instance to local machine:
scp -i "demo.pem" [email protected]:~/CarND-Semantic-Segmentation/runs/1516820578.872475/*.png '/mnt/c/Users/nxa09564/OneDrive - NXP/Udacity/SDCND/projects/term3/CarND-Semantic-Segmentation/runs/1516820578.872475'
- SSH connection to AWS instance:
- Ensure you've passed all the unit tests.
- Ensure you pass all points on the rubric.
- Submit the following in a zip file.
helper.py
main.py
project_tests.py
- Newest inference images from
runs
folder (all images from the most recent run)
- The link for the frozen
VGG16
model is hardcoded intohelper.py
. The model can be found here - The model is not vanilla
VGG16
, but a fully convolutional version, which already contains the 1x1 convolutions to replace the fully connected layers. Please see this forum post for more information. A summary of additional points, follow. - The original FCN-8s was trained in stages. The authors later uploaded a version that was trained all at once to their GitHub repo. The version in the GitHub repo has one important difference: The outputs of pooling layers 3 and 4 are scaled before they are fed into the 1x1 convolutions. As a result, some students have found that the model learns much better with the scaling layers included. The model may not converge substantially faster, but may reach a higher IoU and accuracy.
- When adding l2-regularization, setting a regularizer in the arguments of the
tf.layers
is not enough. Regularization loss terms must be manually added to your loss function. otherwise regularization is not implemented.
If you are unfamiliar with GitHub , Udacity has a brief GitHub tutorial to get you started. Udacity also provides a more detailed free course on git and GitHub.
To learn about REAMDE files and Markdown, Udacity provides a free course on READMEs, as well.
GitHub also provides a tutorial about creating Markdown files.