Technology Demonstrations

Learn about the cutting-edge work we are doing in the areas of computer vision, data and analytics, high-performance computing, scientific visualization, medical computing, and software process.

Detecting Threats

Using a collection of cameras and sensors combined with deep learning, Kitware is developing a system to provide squad commanders an up-to-date assessment of threat conditions.

Reconstructing 3D Buildings from Satellite Data

To aid urban planners, Kitware is developing technology to automatically build 3D models of buildings and other structures from satellite imagery.

Detecting Fake Images and Videos

Kitware is developing media forensics techniques to detect fake or edited images and videos obtained from cell phones, surveillance cameras, satellites, and other sources.

Virtual Training Tools for Laparoscopic Surgeons

Kitware is building a virtual trainer for fundamental laparoscopic surgery that implements core training tasks including suturing, pattern cutting and camera navigation.

Scientific Visualization on Mobile Devices

To handle both the enormous size of simulation data and the need for instant access to the data from anywhere, Kitware’s visualization tools scale from mobile devices to supercomputers.

Software Process

Millions of copies of Kitware’s CMake are downloaded each year by software developers who use the tool to manage the building and testing of their software across multiple platforms.

Detecting Threats

As part of the DARPA Squad X Core Technologies program, Kitware developed THREAT X, a software system designed to be deployed on field conditions to provide squad commanders an up-to-date picture current threat conditions. Unblinking and untiring, THREAT X deploys a collection of color and infrared cameras, along with other sensors, and uses the latest deep learning techniques to identify scene elements (for example, windows and doors), track motion, identify elements that look like people, and identify weapons out to distances as great as one kilometer. THREAT X then intelligently fuses this information into a threat assessment. For example, motion in a door or window could be considered a threat as could an armed person. This assessment is reported to the squad leader for evaluation and, if appropriate, action.

Virtual Training Tools for Laparoscopic Surgeons

Laparoscopic surgery is a type of minimally invasive surgery done through small ports in the body. Long tools and a camera are inserted into the ports. It is a minimally invasive procedure where the surgeon must guide the tools based on the camera feed displayed on a video monitor. This procedure gained prominence (7.5 million per year worldwide) as it avoids causing large open wounds or incisions, thus decreasing blood loss, pain, and discomfort (short term stay in the hospital). However, the procedure requires significant training in the skills required to use the tools. Skills including bimanual dexterity, hand-eye coordination, spatial awareness, and depth perception are necessary for successful laparoscopic surgery.

Currently, surgeons train basic laparoscopic skills using the physical box trainer, FLS box trainer. FLS stands for Fundamentals of Laparoscopic Surgery, which was designed for surgical residents, fellows, and practicing physicians to learn and practice laparoscopic skills to have the opportunity to definitely measure and document those skills. The FLS program has been mandated and endorsed by the American College of Surgeons (ACS) and is a joint educational offering of the Society of American Gastrointestinal and Endoscopic Surgeons (SAGES) and the ACS. This training method requires monitoring by a proctor, as well as the continuous purchase of consumable supplies. Plus, it does not provide objective feedback metrics (varies from proctor to proctor). SAGES has about 80 FLS test centers across US and Canada and is expanding to countries like Israel, France, Singapore, China, and Lebanon.

Our virtual task trainer for fundamental laparoscopy training, designed and tested in collaboration with RPI, Buffalo, and UCA, implements core training tasks including suturing, pattern cutting, and camera navigation. No proctor is required, and no supplies are consumed. Subject performance is automatically evaluated and presented to them at the end of the task.

Reconstructing 3D Buildings from Satellite Data

Kitware is enabling urban planners by providing ways to create up-to-date 3D reconstructions of buildings of any area on the globe from satellite data. Flat, two-dimensional satellite imagery is processed into 3D structures, complete with accurate images on all sides. This will help urban planners from around the world build their own 3D models from the most recent data without relying on Google or Apple to update their 3D buildings provide images for a certain area.

Additional resources:

Scientific Visualization on
Mobile Devices

Numerical simulations have a large impact on our lives though many of us don’t realize their importance. They are used to create new pharmaceuticals to combat disease; investigate new surgical procedures; design new fuel-efficient cars; track and predict changes in our climate and biosphere; develop lighter, stronger materials; and explore new ways of generating energy.

A single simulation can generate over 1,000 times more data than all the written information stored in the Library of Congress and 500 times more data than the content of all of the tweets that have ever been sent since Twitter was founded.

The Scientific Computing Team at Kitware is developing the tools and capabilities to help scientists and engineers analyze and understand simulation results using our award winning software such as the Visualization Toolkit (VTK) and ParaView, as well as specialized tools such as Tomviz, which is used for analyzing and visualizing 3D tomographic information produced by transmission electron microscopy scans of material down to the atomic scale, and Computational Model Builder (CMB), which aids in both defining the simulation and processing the results. Our aim is to support mobile devices; web-technology; and desktop, cloud computing, and super computing environments.

To look to the future, our team is working alongside engineers, scientists, and academic researchers to support the next stage in high-performance computing that will be orders of magnitude more powerful than what is available today.

To give you a taste of what our technology can do, we have prepared a couple demos that can run on your mobile devices using VTK.js and ParaView Glance.

Flow Around a Nuclear Reactor Pin

CT Scan of an Upper Torso

X-Ray of an Aneurysm

X-Ray of a Tooth

Lysozeme – Enzyme in both tears and egg whites

T-Bar Assembly Part

T-Handle Flange Part

More Paraview Glance Demos

Detecting Fake Images

Modified and fake images and videos have become a serious challenge for validating and authenticating visual media content in news stories, social media, and other sources. Perpetrators can easily change a region of a real image or video to offer a very different narrative, as seen in the video related to the Parkland shooting where the rifle target being torn in half by the student was replaced with the US Constitution. Computer algorithms can also generate new fake images or videos with approaches like deep learning. Some of these fake images have become so realistic that many people cannot distinguish them from real images. Furthermore, Deepfake videos of high profile political figures like the US president and other celebrities can be very damaging, purportedly showing politicians saying words they did not say, and have received a lot of attention in the media lately.

Under the DARPA Media Forensics program, Kitware is developing technology to detect manipulated or fake visual media, leading a team of five research universities including Columbia University, Dartmouth College, the University at Albany, the University of California at Berkeley, and the University of Oregon. We are developing a variety of media forensics techniques to detect fake or photoshopped images and videos, including deep learning algorithms that can also localize manipulated parts of the images or videos. In 2018, our holistic manipulation detection algorithm performed the best of dozens of entries in the Media Forensics Challenge conducted by NIST and DARPA. Our algorithms are applicable to images and videos from a variety of sources like cell phones, surveillance cameras, satellites, and others.

Software Process

Kitware builds and maintains a suite of open source solutions for use in the software development process. These solutions facilitate high-­quality software development by providing a cross­-platform tool to build and compile software, a software testing framework to evaluate new code contributions, and a dashboard to highlight test results and overall software health. We use our software development process internally and work with external customers to facilitate iterative processes that ensure consistency and quality in finished software products.

About Kitware

Kitware is Hiring

Learn About Our Expertise

Our Projects & Technology