Philippe-Henri Gosselin

Full Professor at ENSEA, ETIS Lab., Cergy, France
Team : Midi, Office : D383
Address : ENSEA,
6 avenue du Ponceau, 95000 Cergy
Tel : +33 1 30 73 66 11, Fax : +33 1 30 73 66 27
Email : philippe-henri.gosselin@ensea.fr
Current thematics

Decentralized learning for Multimedia Indexing and Retrieval

decentralized

Machine learning in decentralized environment using Gossip protocols:

  • Randomly distributed data in networks or computing devices
  • No central entity
  • Full asynchronism

Proposed Gossip methods:

Supervisor of Ph.D. student : Jérôme Fellus

Tensor-based Representations for Multimedia Indexing and Retrieval

matching

Linearisation of visual descriptor matching techniques

Applications:

  • Near-duplicate image search in large databases (1 million)
  • Image categorisation (VOC2007)
  • Video categorisation (KTH, Hollywood2)

Available demonstration software on http://www.vlat.fr

Supervisor of Ph.D. student : Romain Negrel

Indexing and Machine Learning for Remote Sensing Images

RSI

Multi-Scale Classification of Remote Sensing Images (RSI)

Supervisor of Ph.D. student : Jefersson Alex dos Santos

Past thematics

Kernels on Bags and Graphs for Image Retrieval

graph

Kernels functions on visual components

Applications:

Supervisor of Ph.D. students : Justine Lebrun and Jean-Emmanuel Haugeard

Interactive and Collaborative Learning for Content-Based Image Retrieval

retin

Overview in French language: Ph.D. and HDR

Interactive Learning:

Collaborative Learning:

Supervisor of Ph.D. student : Alexis Lechervy

XHTML 1.1