SPARSIS

Sparse Modeling and Estimation of Motion Fields

 

 

 

 

 

 

 

 


Home
Administrative info

Research team

Summary
Tasks
Publications
Consultants

Awards

Protected Area



1.    Management and Dissemination

This task aims at controlling the project execution so that the work flow proceeds according to the schedule.

 

2.    Sparse Estimation of Multiple Motion

First we have to devise efficient methods to estimate the model parameters. Since the number of parameters is very high (many

hundreds) efficient representations are required. This task explores the use of sparse representations of the motion fields and switching matrices.

 

3.    Online Motion Field Estimation

Previous research (ARGUS project) dealt with offline learning of motion fields. In this project, we propose online learning to overcome both difficulties. This will be achieved in two main stages:  i) Single motion field estimation;

ii) Multiple motion field estimation

 

4.    Multicamera Systems

In this task we will tackle the extension of the motionfield paradigm to the modeling of multiple camera systems and exploit sparse methods for learning their topology and intercamera spatiotemporal relations. Learning will be performed in datasets of distributed camera networks:  i) Motion fields in Multicamera Systems; ii) Learning Network Topology; iii) Cameras with overlapping fields of view; iv) Cameras with nonoverlapping

 

5.    Activity Recognition 01012018

This task will embrace the following goals: i) Development of online inference algorithms for classifying activities in multicamera systems according to the learned models;  ii) Development of methods to cope with persons' interactions;  iii) Evaluate the impact of sparsity in activity recognition. We will compare different activity recognition and model learning methods, with and without sparsity constraints.

 

6.    Nonlinear System Identification with Sparse Models

This task aims at exploiting the interplay between methods that lead to sparse models for nonlinear system identification and the

ones considered in the project for the analysis of video sequences, relying on multiple models and sparse estimation.

 

7.    Final Report

In this task, the final report is that reports a synthesis of the main results obtained in the project is produced and a synthesis of

the work produced during the project that yielded them

 

 

 

Task Nº

Task Denomination

Participant responsible for task

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

1

Management and Dissemination

Jorge Marques

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2

Sparse Estimation of MMF

Mário Figueiredo

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3

Online MF Estimation

Miguel Barão

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4

Multicamera Systems

Alexandre Bernardino

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5

Activity Recognition

Jacinto Nascimento

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6

Nonlinear Identification of SS

J. Miranda Lemos

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

7

Final report

J. Miranda Lemos