Skip to content

vpulab/wisemnet

Repository files navigation

WiSE-MNet: Wireless Simulation Environment for Multimedia Networks.

This repository contains the Wireless Simulation Environment for Multimedia Networks (WiSE-MNet).

WiSE-MNet is based on Castalia/Omnet++ and enables the modeling of the communication layers, the sensing and distributed applications of Wireless multimedia sensor networks (WMSNs), i.e. networks with sensors capturing complex vectorial data, such as video and audio. The environment is designed to be flexible and extensible, and has a simple perception model that enables the simulation of distributed computer-vision algorithms at a high-level of abstraction.

The main features are:

  • Standard operations: sensing, processing, communication, and decision making
  • Wireless and Wired communication support
  • Templates for fast prototyping
  • Power-consumption models for smart-camera hardware
  • Simulates realistic multi-camera networks with both realworld and synthetic datasets

Please cite the following reference when using the simulator or any included file:

Networked Computer Vision: the importance of a holistic simulator, 
J. SanMiguel, A Cavallaro
IEEE Computer, 50 (7): pp. 35-43, July 2017

Authors

Current version:

Original version (WiseMnet 2011):

Distribution as a virtual machine

To avoid installation issues, a Virtual Machine (VM) is provided to ease the development of applications in Wise-Mnet. It can be downloaded [here] (5.3GB, 17GB uncompressed) and a tutorial to use the VM is available [here]

The VM runs using the VirtualBox manager which is required to be installed in your host system [Linux download] and [Windows download]. Thus, you can run WiSE-MNet from Windows or Linux via VirtualBox.

The provided VM has been created with minimal capabilities with the following main details:

  • OS Ubuntu 16.04 LTS
  • 2 CPU
  • 4GB RAM
  • Hard Drive (max 40GB, current usage 17GB)
  • Software installed: required libraries for WiseMnet v1.4 (Omnet++ 5.0 and OpenCV 3.2.0)
  • User=wise password=wisemnet

Once the VM is added to the VirtualBox manager, we suggest to change the VM features (i.e. increase the number of CPUs, RAM and HDD space) according to your system capabilities.

Documentation

The documentation of the code is available at the file doc/UserGuide-v1.4.pdf. Additionally, documentation using doxygen style is provided in the directory 'doc/html' (see 'doc/html/index.html' for more information).

Moreover, we provide tutorials via Youtube [list]:

  • WiseMnet++ tutorial: install from github using command line [video]
  • WiseMnet++ tutorial: install from github using GUI-Omnet++ 5.0 [video]
  • WiseMnet++ tutorial: changing settings for a simulation [video]
  • WiseMnet++ tutorial: downloading and using multi-camera datasets [video]

Required libraries and installation

Wise-MNet requires the following libraries:

  • OMNeT++ version 5.0
  • Castalia version 3
  • OpenCV version 3.2.0
  • Eigen3 version 3.2.92

Two installation scripts are provided install_omnetpp.sh and install_opencv.sh. Further installation instructions are available at doc/UserGuide-v1.4.pdf

Dataset downloading

Currently, WiseMnet++ employs the following datasets:

  • Sequence S2_L1 from Crowd_PETS09 dataset (7 views) [link]
  • Sequence AB_HARD from AVSS2007 dataset (1 view) [link]
  • Sequence CHAP from ICGLab6 dataset (4 view) [link]

Sample bash scripts are provided to download the video frames and the associated data (groundtruth, calibration,...) in the directory ./videos. These scripts prepare all data to be used directly in Wise-Mnet++.

Available algorithms in the simulator

The list of available algorithms is given in the following table:

Algorithm Type Focus Approach Reference
KCF: Kalman Consensus Filtering Tracker Multi-camera single-target Consensus-based IEEE CDC 2009 [link]
DPF: Distributed Particle Filter Tracker Multi-camera single-target Token-passing IEEE SSPD 2011 [link]
ICF: Information Consensus Filtering Tracker Multi-camera single-target Consensus-based IEEE TAC 2013 [link]
CAWC: Cost-AWare Collaborative tracking Tracker Multi-camera single-target Consensus-based IEEE SENSORS J 2015 [link]
MTIC: Multi-Target Information Consensus Tracker Multi-camera multi-target Consensus-based IEEE TPAMI 2016 [link]
People detect&transmit Detector Single-camera - IEEE COMPUTER 2017 [preprint]

References using the simulator

Currently, the simulator has been used in the following papers:

  • WiSE-MNet: an experimental environment for Wireless Multimedia Sensor Networks, SSPD 2011, [preprint]

  • Distributed target tracking under realistic network conditions, SSPD 2011, [preprint]

  • Cost-aware coalitions for collaborative tracking in resource-constrained camera networks, IEEE Sen. Journal 2015 [link],

  • Energy Consumption Models for Smart-Camera Networks, IEEE TCSVT 2016 [link],

  • Networked Computer Vision: the importance of a holistic simulator, IEEE COMPUTER 2017 [preprint]

License

The simulator code and the documentation is available under the Academic Public License (same as OMNeT++ and Castalia). See the "LICENSE_WISE" for more information.