Semantic segmentation and depth estimation with hardware restricted fully ConvNets (FCNs) for autnomous driving. Please proceed with the jupyter-notebook "Brief_Results.ipynb"
This repository should do two things: 1) show that it is possible to do dense prediction tasks with limited hardware on huge state-of-the-art datasets and get utilizable reults. 2) provide code examples that are kept easy to understand and meant to be re-used for different projects involving FCNs in TensorFlow.
To get started you need: 1) TensorFlow V.1.4+ with python 3 and GPU support, 2) A GPU with at least 4 Gbyte RAM, 3) Some basic understanding of python and convolutional neural networks (or better FCNs)
If you lack the basics mentioned above I suggest you to read >> http://neuralnetworksanddeeplearning.com >> https://arxiv.org/pdf/1605.06211.pdf >> http://www.deeplearningbook.org/
Please visit >> https://github.com/tensorflow/models/tree/master/research/slim