The pixels of a road in images using a Fully Convolutional Network (FCN). The FCN is trained on KITTI data set.
The intelligence in the pre-trained network VGG-16 is used in the FCN by replacing the fully connected layer with a 1 X 1 convolution in order to preserve the spatial information. The 1 X 1 convolution output is upsampled using Transpose Convolution to construct the output. Skip Connections are used to prevent the degradation problem.
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.
Run the following command to run the project:
python main.py
The results can be found in the runs folder