IPP-UPSAY

Master 2 Course

"Image Mining"

Schedule 2024-2025: November-December / Wednesdays morning

Location: Télécom Paris, Palaiseau (Map)

Lecturers: Antoine Manzanera (ENSTA / U2IS), Gianni Franchi (ENSTA / U2IS), Flora Weissgerber (ONERA)

Objectives, Pre-requisites and Content
 
Images and Videos represent a major source of information today.
There is an increasing demand for automated methods to organize huge collections of image data, and to interpret images and videos by computer.

The Image  Mining  course  deals  with  the  problem  of  increasing  the semantics of visual data by: (1) reducing its information to relevant data, and (2) indexing it relatively to a model and/or to a previous knowledge, in order to facilitate its retrieval and interpretation by a machine.

The  course  does  not  assume  previous  knowledge  in  Image  Processing,  but basics  in  signal  processing,  information  theory  and  pattern  recognition  are useful.

After presenting the different mathematical models for images, the feature extraction and description methods are described.
One third of the course is dedicated to Classification and Machine Learning methods : Clustering, Dimension Reduction, Bayesian learning, Neural Networks.
The video analysis problem is addressed through the fundamental task of Visual Object Tracking.
An important case study is extensively presented: the mining and retrieval of Remote Sensing images.
Evaluation Mode

Grade = 0.5 * Practical works + 0.5 * Written exam
  • 1 practical work to be chosen between Bag-of-Features Mining and Visual Tracking (w = 1)
  • 1 practical work on Remote Sensing (w = 1)
  • 1 written exam on the whole program (w = 2)

  TIME 
ROOM SESSION DESCRIPTION
 LECTURER RESOURCES
Nov, Wed. 13
9:00 am		 1C43
  • Introduction: Image models and fundamental tools
 Antoine Manzanera
Nov, Wed. 20
9:00 am		 1C43
  • Image representation: Image feature extraction, Local and Global descriptors, Image Indexing and Search
  • Image classification and Visual Learning: Dimension reduction, Clustering methods, Bayesian learning, Other supervised and unsupervised methods
  • Practical work #1: Bag-of-Features based Image Mining
 Antoine Manzanera
Nov, Wed. 27
9:00 am		 1C43
  • Deep learning for image mining: Neural Networks, CNN, Deep NN training, Regularisation, GAN
 Gianni Franchi
Dec, Wed. 4
9:00 am		 1C43
  • Visual Tracking: Local methods, Distribution based methods, Hough transforms, Predictive filtering
  • Practical work #2: Object Tracking
 Antoine Manzanera
Dec, Wed. 11
9:00 am		 1C27
  • Application / Case study: Earth Observation and Remote Sensing Image Mining
 Flora Weissgerber
Dec, Wed. 18
9:00 am		 1C27
  • Practical work #3: Case study on Remote Sensing
 Flora Weissgerber
Jan, Wed. 8
9:00 am		 0C03
  • Written exam (~1h)
  • Q/A session on the practical works (~2h)
Antoine Manzanera