Description
The aim of this paper is to introduce to the newcomers the ideas of Deep Neural Networks started by Yan LeCun and continued by Alex A., NYU, Google and Facebook teams, make a small panorama of the more common types of Neural Networks available and explain in detail a new and very successful architecture called Mask R-CNN that has won recognition all around the world.
After this big introduction, we will dive into the resolution of the problem of Lane Recognition with images taken from inside cars using CuLanes dataset and its implementation in TensorFlow. We will see how difficult and problematic this type of images can be due to the different and possible geometric issues that diverse landscapes have. Nevertheless, we will show that the technique is applicable to this specific problem and could be improved to be automatized and implemented in a self-driving car.
https://github.com/fmcalcagno/MASK_Lane_Detection @fmcalcagno