geraudmm Goto Github PK

advanced-lane-finding

In this project, the objective is to use the OpenCV library to perform track line recognition. We will start by testing our algorithm on images and then directly on videos.

attractiondistancegame

In this project, we use NeuroEvolution to train 2 different agents. The first one wants to reach the second one who is trying to escape.

behavioral-cloning-for-a-self-driving-car

In this project, we'll train a Convolutional Neural Network to drive a car, using data made by a human player in a simulation.

build-ml-pipeline-for-short-term-rental-prices

Project code for cd0581 refresh taught by Giacomo Vianello

carnd-path-planning-project

In this project, the goal is to design a path planner that is able to create smooth, safe paths for the car to follow along a 3 lane highway with traffic.

carnd-pid-control-project

In this project we'll implement a PID controller in C++ to maneuver a vehicle around the track from the Behavioral Cloning Project!

clustering-dynamique-pour-la-d-tection-et-le-suivi-d-objets-en-perception-interactive

We propose a method aiming at giving the possibility to a robot, to understand by interacting with its environment, which parts belong to moving objects and which ones belong to the background. The originality of this approach is the use of very few hypotheses in order to allow better generalization in the most diverse environments. The novelty comes from the use of a persistent segmentation of the scene giving the possibility to follow each part of the scene and to better understand the boundary of each object.\\ The benefit of this method is that it provides a solid basis for using an affordance map to understand which actions are possible for each part of the environment. Thus, once the robot is able to understand how and where to interact with its environment, it will be easier to make it perform a more complex set of tasks.

collab_compet-with-d.r.l.

In this project, we are teaching a DDQN agent to play tennis against himself in a Unity environment.

continuous-control-with-deeprl

In this project, the goal is to implement a deep reinforcement learning algorithm that can deal with continuous action space.

deploying-a-scalable-ml-pipeline-in-production

Deploying a ML Model to Cloud Application Platform with FastAPI

extended-kalman-filter

In this project, we will use an extended kalman filter relying on a Radar and a Lidar to predict the next position and velocity of a vehicule on a 2D map.

finding-lane-lines

In this project, the objective is to use the OpenCV library to perform track line recognition. We will start by testing our algorithm on images and then directly on videos.

models

Models and examples built with TensorFlow

navigation-with-drl-in-a-unity-environment

The goal of this project is to train an agent to collect object in a 3D world. we'll use a Deep Q-Network in a discrete action-space.

neuroevolution-on-flappy-bird

In this project we train multiple neural networks to play flappy bird. At each generation we choose the Agents that performed the better and mutate them a little in order to converge to an optimum neural network.

p9---capstone-

particle-filter

In this project, we will try to obtain an accurate estimation of the location of a moving car through the use of a particulate filter.

real-pr2-clustering-dynamique-pour-la-d-tection-et-le-suivi-d-objets-en-perception-interactive

We propose a method aiming at giving the possibility to a robot, to understand by interacting with its environment, which parts belong to moving objects and which ones belong to the background. The originality of this approach is the use of very few hypotheses in order to allow better generalization in the most diverse environments. The novelty comes from the use of a persistent segmentation of the scene giving the possibility to follow each part of the scene and to better understand the boundary of each object.\\ The benefit of this method is that it provides a solid basis for using an affordance map to understand which actions are possible for each part of the environment. Thus, once the robot is able to understand how and where to interact with its environment, it will be easier to make it perform a more complex set of tasks.

render-cloud-example

tetrapode-robot-drl

In this project we will learn to the robot to walk in a Unity simulation using the PPO(Proximal Policy Optimization) algorithm for Deep Reinforcement Learning

traffic-sign-classifier

In this Project, we will try to sort the German Traffic Signs using a Convolutional Neural Network.