Hi! I'm Thomas Robert, PhD Student in Deep Learning for Computer Vision.

I started my PhD on November 2016, and I am supervised by Matthieu Cord and Nicolas Thome at the LIP6 (UPMC Paris 6 / Sorbonne Universités)

Research

Our team

I am working at the Pierre and Marie Curie University in Paris, at the LIP6 lab, in the MLIA team managed by Patrick Gallinari. Our team, supervised by Matthieu Cord, is focusing on Computer Vision research and is of course taking the deep learning turn. Here are some example of publications by our team:

  • WILDCAT: Weakly Supervised Learning of Deep ConvNets for Image Classification, Pointwise Localization and Segmentation. T Durand, T Mordan, N Thome, M Cord. CVPR 2017.
  • LR-CNN for Fine-grained Classification with Varying Resolution., M Chevalier, N Thome, M Cord, J Fournier, G Henaff, E Dusch. ICIP 2015.
  • Recipe Recognition with Large Multimodal Food Dataset. X Wang, D Kumar, N Thome, M Cord, F Precioso ICMEW 2015.
  • MANTRA: Minimum Maximum Latent Structural SVM for Image Classification and Ranking. T Durand, N Thome, M Cord. ICCV 2015.
  • Sequentially Generated Instance-Dependent Image Representations for Classification. G Dulac-Arnold, L Denoyer, N Thome, M Cord, P Gallinari. ICLR 2014.
  • Fantope Regularization in Metric Learning. MT Law, N Thome, M Cord. CVPR 2014.
  • Learning deep hierarchical visual feature coding. H Goh, N Thome, M Cord, JH Lim. TNNLS 2014.
  • Top-Down Regularization of Deep Belief Networks. H Goh, N Thome, M Cord, JH Lim. NIPS 2013.
  • Pooling in image representation: The visual codeword point of view. S Avila, N Thome, M Cord, E Valle, ADA Araújo. CVIU 2013.
  • Quadruplet-wise Image Similarity Learning. M Law, N Thome, M Cord. ICCV 2013.
  • Unsupervised and Supervised Visual Codes with Restricted Boltzmann Machines. H Goh, N Thome, M Cord, JH Lim. ECCV 2012.

Research context

I am working on the DeepVision project, which is a research partnership between French and Canadian teams funded by the ANR (France) and NSERC (Canada). This project involves the LIP6 / UPMC (France) with Matthieu Cord, the LIRIS / INSA Lyon (France) with Christian Wolf, the Simon Fraser University (Canada) with Greg Mori and the University of Guelph (Canada) with Graham Taylor.

Research interests

My research interests are broad and goes from supervised deep learning for image classification to semantic representation learning for text and images and deep recurrent neural networks.

Currently, I am working on developing new architecture of convolutional neural networks for image classification using semi-supervised learning, and for this matter I have a particular interest on generative models of images (VAEs, GANs, etc.).

Recent work

M2CAI workflow challenge: classification of surgical operation steps based on endosopic videos

R. Cadene, T. Robert, N. Thome, M. Cord

M2CAI workshop @ MICCAI (2016) 2nd best model submitted to the challenge

Technical paper Slides

VISIIR: deep learning for recipe image classification

T. Robert, R. Cadene, N. Thome, M. Cord

(2016) Unpublised extension of published work

Project website & Demo Related publication

DjLu: an open-source tool to organize your bibliography

T. Robert, M. Carvalho, R. Cadene

GitHub (2016)

Website Code

Teaching

During my PhD, I am teaching assistant on various classes at the University Pierre and Marie Curie (UPMC). I am teaching:

Java and Object-Oriented Programming

2I002 Course, L2 students (50h)

Course page

Neural Networks and Deep Learning for Pattern Recognition

RDFIA Course, M2 students (10h)

Course page

Practical Introduction to Deep Learning for Image Classification

Multimedia Course, Polytech M2 students (8h)

Course page

Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA)

BIMA Course, M1 students (8h)

Course page Data for practical session