Development of a strategy to augment a data set of medical images of the lower
limb to train a Neural Network for the automatic segmentation of bones out of
CT-scan images.
The final application for the company is the simulation of the installation and
functionality of a knee prosthesis adapted to specific patients.
Developed skills: medical imaging basics, 3D images & meshes
processing, statistics (Principal component analysis), and Deep Learning.
Adaptation of a 2D inpainting
Convolutional Neural Network (CNN) to 3D images (Groups of 3)
Working with researchers of Jean Kuntzmann Laboratory (Grenoble) on
protein’s 3D structure prediction using deep learning.
Use of Python’s libraries TensorFlow and PyTorch.
We modified an existing model to be able to reconstruct 3D images with
masked regions (beyond the state of the art).
Creation of an animated scene of the Middle Ages with 3D objects.
Use of computer vision technics together with mathematical tools (eg.
illumination, rotation, animation).
Simulation and restoration of fingerprint acquisition problems (blurry,
pressure, rotation, missing parts…)
Use of OpenCV library for image loading/writing.
Implementation of mathematical notions: convolution, rotation,
interpolations.