Tackle challenging image and video analysis problems with KWIVER

Kitware offers advanced R&D services in support of KWIVER. Find out how we can help develop an end-to-end solution for your challenges in automated image and video analysis.

The Kitware Image and Video Exploitation and Retrieval (KWIVER) toolkit is a collection of software tools designed to tackle challenging image and video analysis problems and other related challenges. Recently started by Kitware’s Computer Vision and Scientific Visualization teams, KWIVER is an ongoing effort to transition technology developed over multiple years to the open source domain to further research, collaboration, and product development.

Existing KWIVER Open Source Repositories

KWIVER: The main repository supporting the rest of the KWIVER ecosystem.  The KWIVER “vital” libraries provide core data structures, algorithms APIs, logging, and configuration infrastructure useful for building your own exploitation systems.  The KWIVER “arrows” provide algorithm plugins from various third-party open source packages so you can dynamically add on the functionality that you need. KWIVER includes Sprokit, the “Stream Processing Toolkit”, a library aiming to make processing a stream of data with KWIVER arrows easy. It supports divergent and convergent data flows with synchronization between them, connection type checking, all with full, first-class Python bindings. This repository is where you should start.

Fletch : A CMake based project that assists with acquiring and building the common Open Source libraries that KWIVER uses in its collection of arrows.  If you want to build KWIVER with all the bells and whistles, Fletch makes it easy to get the right versions of everything you need and makes sure they all integrate together.  You get the same versions of packages working together whether you are on Windows, Linux, or Mac. Fletch provides everything you need from libjpeg, libjson, and ffmpeg to Qt, VTK, OpenCV, and more.

TeleSculptor TeleSculptor is an open source, end-user desktop application for aerial photogrammetry. It leverages KWIVER algorithms to build 3D scene models from aerial video, such as video collected from UAVs. It handles geospatial coordinates and can make use of metadata, if available, from GPS and IMU sensors. However, it can also work with non-geospatial data and with collections of images without associated metadata. Although it is primarily an end-user application, it is highly reconfigurable to also support photogrammetry research and development.

To download the latest version of TeleSculptor, use one of the following links based on your operating system:

Instructions on how to use TeleSculptor can be found in the User GuideKitware offers advanced software R&D solutions and services. Contact us to discuss how we can help with your next TeleSculptor project.

Social Multimedia Query Toolkit (SMQTK) : A collection of Python tools, with C++ dependencies, for ingesting images and video from social media (e.g. YouTube, Twitter), computing content-based features, indexing the media based on the content descriptors, querying for similar content, and building user-defined searches via an interactive query refinement (IQR) process.

VIVIA : A collection of Qt-based applications for GUIs, visualization, and exploration of content extracted from a video.

KWant : A lightweight toolkit for computing detection and tracking metrics on a variety of video data. It computes spatial and temporal associations between datasets, even with different frame rates.

Existing KWIVER Government Open Source Repositories

Kitware’s WAMI Tracker :The product of multiple years of development, Kitware’s real-time, full-frame WAMI tracker has been used operationally overseas. The tracker is available as source code for government use. Access is available to government entities and to those with government sponsorship.  To inquire or receive source code, please reach out to computervision@kitware.com.

Existing Data Sets

VIRAT video dataset : The VIRAT video dataset contains more than ten hours of realistic, natural, and challenging video surveillance data, particularly in terms of its resolution, background clutter, diversity in scenes, and human activity/event categories. Ground truth is provided for all movers and more than 10 event types.

Multiview Extended Video with Activities (MEVA) Dataset:The MEVA dataset aims to build a corpus of activity video collected from multiple viewpoints in realistic settings. The first release, “KF1” (Known Facility Release 1), contains 100 actors across 29 cameras for a total of 185 hours of video.

Planned Public Open Source Repositories & Data Sets

Kitware is working toward making the following repositories and data sets available:

  • WorkQL : A PostgreSQL based synchronization library for developing distributed, pipeline- oriented, data processing systems.