For the Transversal Project of the master VAR, we had to create an AR or VR application by groups of two. My colleague and I decided to realize an AR application for a street art in Evry. This fresco was created by children of the “Maison de Quartier”, students of Telecom Sudparis and the street artist Vince. We wanted to make an application in connection with art, and we knew the vast number of frescos in Evry.

For the AR class I created an MRTK application on Unity. The goal of the app is the following : “We are in 2027. You are the CEO of an AR startup. Today is the Paris Motor Show. You had the chance to make a deal with the organizers of the event. You are demonstrating your AR application to the public. A good review of the visitors will get you a big fat contract with the Paris Motor Show.

The goal of this lab is to reconstruct a 3D object using 2D images, with the C3DC software. First, I chose to reconstruct the following ladybird. So I put her in a room with homogenous lighting, and took images of it from various angles. However, the images cannot be unconnected, the algorithm was not able to match the feature points on the object and its background. Therefore I added a white sheet of paper beneath the ladybird, as you can see on the following images :

This post is the second part of my evaluation for the Recurrent Neural Network class. I am going detail what I did during the laboratory. The lab consisted in implementing a RNN for image captioning, you can find the code on the Google Collab here : https://colab.research.google.com/drive/1qft4rTOrA_FAP6kAPhAMfDisNmcfUh-r?usp=sharing I completed it with all that we have seen during the class, and I tried to train the model by my own. As an input of the model , I must give images and their partial caption input, and as the expected output the words to predict.

I followed a class about Recurrent Neural Networks. In this first part I will explain what I learned from it, and in a second part in the next post I will explain the lab I did on the same topic. (All the images used come from my teacher’s presentation.) Presentation A Recurrent Neural Network (RNN) can, as any neural network, approximate a non-linear function by learning from the given data, in order to output the required vectors.

This post is the second part of my work for the Reinforcement Learning labaratory. I presented the reinforcement learning model in the previous post, and here I am going to explain what I changed in the model to see the impact of the different parameters. Part 2 : Manipulations First test I conducted a first training with the basic following parameters : And I looked at the first results without changing any of those parameters :

In this post I will explain one of the laboratories I did this year about Reinforcement Learning. During this lab, we used a python code on Google Collab to manipulate a deep reinforcement learning network to solve the CartPole game : a game where a hammer must not fall. In this first part, I am going to explain the different functions of the model, and in the second part in the next post, I will explain what I changed in the model to see the impact of the different parameters.

In my Advanced Computer Vision class, I got a lab about the 3D reconstruction of human poses. The purpose is to The purpose of the lab is to implement the 3D reconstruction from multi-views of human pose on a calibrated system. We were given various files, such as videos of each subjects, and the visual output of the reconstruction on each sequence of each subject. I started by trying to run the reconstruction code given on the sequences of the subject Lea, to see what we were supposed to obtain :

Hi ! In this post I describe how I made a simple AR application with image recognition. I started by looking at tutorials on how to use AR on Unity. First, I followed the tutorial on Unity learn, to set up my project for AR : https://learn.unity.com/tutorial/configuring-unity-for-ar-development-2019-3?language=en#5e623ad9edbc2a0020cf8224 And then this video, because it talked about Image recognition an aspect we will need for our project : https://www.youtube.com/watch?v=9qErhhxoEVY Si I started to put in place all the components needed for an AR project shown on Unity Learn.

Hello everyone ! Today, I created my first VR application for Oculus Quest. I modified my Roll a Ball game to work in VR. First, I downloaded the VRTK project from VRTK4’s Github, in order to use some of its packages. I changed the project settings as followed : Then, still in the VRTK project, I imported the Oculus Integration package. I exported my Roll a ball project assets, and imported it in my VRTK project :

Hi everyone ! During the fourth HCI lecture, we focused on mobiles and the concept of Ubiquitous Computing. Smartphones Why did they change their shape and what do you think will be the next step in this transformation ? Explain why you think this could be the case. As we can see on this image from the HCI lesson, from the 1990 to 2000 smartphones shrunken along the years. They became smaller and slimer in order to fit in our pockets, and be taken everywhere.

Hello everyone! For the Unity class, I made my own Platformer Game, following Lea’s lessons and some Youtube tutorials. Moving Platforms First, I learn how to create a moving platform. I created a platform, with a RigidBody and a Box Collider. In order to make it move, I first used an animator, and defined its movement. However, even with a velocity script, my character could not move once he was on the platform.