|
SPARSIS Sparse
Modeling and Estimation of Motion Fields |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
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
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
