It is an interesting time to be a data scientist. Tools, techniques, and expectations in different sub-genres are evolving asymmetrically. Most notably, I feel that visualization has lagged machine learning for quite a long time. I must clarify: While colorful corporation-friendly visualization platforms are thriving, technical visualization tools used by data scientists remain stagnate. By technical visualization tools, I am referring to the bare-bones developer-friendly tools that let data scientists peer into large data sets to better understand them. Some of these include the native plot() and the library ggplot2 in the R Language, as well as matplotlib for Python.
As I mentioned in this post on Visualizing N-Dimensional Data, I am less than pleased with the R Language’s native ability to help us understand complex data. Some things as simple as the famous iris, wine, or mtcars data sets are still hard to wrap my head around even though they only have less than 200 rows.
After experimenting with Blender’s Python API and its builtin Virtual Reality (VR) support, I realized the open-source community was already very close to VR-enabled data visualization. The concept of viewing data via a Microsoft Hololens, Google Cardboard, or an HTC Vive was always possible, but seemed very out of reach. The types of professionals that regularly deal with 3D modelling are artists, special effects experts, and game developers. Data scientists tend to dabble in 3D visualization when the need arises but never really attack the problem at a low level. Generally, these types of minds don’t typically hang out together. There is rarely a professional purpose to do so.
Thankfully, the advent of scriptable 3D modelling (such as in Blender’s Python API) has allowed us to cut game developers out of the equation. One data scientist, with the aid of scriptable 3D modelling, and one Android developer, armed with a Google Daydream, can take multivariate visualization to the moon.
See the unveiling of our VR-enabled data visualization software April 13th at 5:00 at University of Virginia, Nau 101.
Could your data visualization software tie in with other common operating platforms, such as Priority 5’s TACCS system? Right now Priority 5’s Touch Assisted Command and Control System (TACCS™) provides a net-centric, service oriented architecture (SOA) geospatial operating platform, built upon open standards, that integrates data, simulation, and tools into a single dynamic environment. Users from remote locations collaborate in real-time through workgroups. Each workgroup brings all the integrated capabilities of the system into a true live COP that goes well beyond the static information available using conventional viewers.
Based on your information in your blog I think an introduction to Priority 5 is a logical next step. I’m happy to connect you to them.