You can freely download the executable files for running DVA, and for visualizing its behavior.
Any comment is precious for the development of our project, so feel free to contact us for any question regarding the software!
Once you have downloaded the software, please send us an e-mail at one of the following addresses in order to get an activation key:

dva AT diism DOT unisi DOT it

DVA (version 3.12 - browse other versions)

The main DVA application is written in C/C++ and requires OpenCV and FFmpeg for simple video handling functionalities.
You can download the installation package for several operating systems and architectures, as well as some scripts for support in dependencies installation (see README file in the downloaded archive):
Linux (gcc version < 4.8)
Linux (gcc version >= 4.8)
MacOSX (up to Lion, clang)
MacOSX (up to Lion, gcc48 from MacPorts)
MacOSX (from Mavericks, clang)

VIEWER (version 2.0 - browse other versions)

The DVA visualizer is written in Java 7. It allows users to monitor experiments both running locally and on remote machines. In the latter case, several communication protocols are implemented, all relying on SSH-based connections to the server machine.

MANUAL (version 1.1 - browse other versions)

A manual describing the DVA software, as well as briefly illustrating the use of the DVA Viewer, is available for download. Please feel free to contact us for any question which is not clear or not directly answered in the manual.

I/O TOOLS (version 1.0 - browse other versions)

A simple package collecting basic MATLAB and C++ routines to load, edit, save data in DVA format.


Here you can find some simple DVA agents that are ready-to-use. We prepared them by processing several hours of misc video data to extract low-level features. These sample models were prepared with the aim of keeping the DVA architecture small and efficient, instead of focussing agents with a large number of features and stronlgy developed invariance properties.
Model 001: transf_unrolled
Model 002: simple_fifteen
Model 003: simple_monocat
Model 004: simple_multicat