[Pycon] [new paper] "Lou Marvin Caraig" - Computer Vision and Machine Learning for Self-Driving Cars with OpenCV and Keras

[info a pycon.it]

Title: Computer Vision and Machine Learning for Self-Driving Cars with OpenCV and Keras
Duration: 90 (includes Q&A)
Q&A Session: 20
Language: en
Type: Talk

Abstract: ## Target audience

The ideal audience for this talk is a passionate person in Machine Learning especially when applied to images and videos. The talk doesn't require advanced knowledge of Python and doesn't require any previous Computer Vision knowledge. Only a basic understanding of supervised learning is required and knowing what a CNN is would be recommended.

## Summary

Did you ever want a self-driving car while in the middle of the traffic? Did you ever want to develop the intelligence controlling that self-driving car? Well... I did. And this talk will show how we can easily develop a Convolutional Neural Network with Keras, train it with a dataset augmented with OpenCV, and also test it on a simulator made with Unity. All the code will be written using Python 3.6.

## Abstract

The talk will be structured as follows:

1. Brief overview regarding self-driving cars
2. Brief introduction about the programs and the libraries that will be used:
   * [OpenCV](https://opencv.org/)
   * [Keras](https://keras.io/)
   * [Unity](https://unity3d.com/)
3. Explanation on how the self-driving car simulator works ([Udacity's Simulator](https://github.com/udacity/self-driving-car-sim)) and how to create the dataset
4. The dataset will be made by images and some values such as throttle, steering angle, etc.. In this step we will talk about the importance of data augmentation by applying many different image transformations using OpenCV such as changing the brightness, adding shadows, adding vertical shifts, etc..
5. Building and training of the CNN with Keras using the architecture introduced by NVIDIA in [this paper](https://arxiv.org/pdf/1604.07316v1.pdf)
6. Explanation on how to attach the CNN to the simulator

At the end of talk we will finally see the self-driving car in action!

Tags: [u'ComputerVision', u'neural network', u'Machine Learning', u'Keras', u'opencv', u'machine-learning', u'Artificial Intelligence']