AEONX: S0L

AEONX: “S0L” The Head Motion Scanning System

IR’s AeonX head motion scanning system called S0L was designed and built from the ground up utilizing Ximea’s machine vision cameras featuring the high quality Sony Pregius sensors. Our cameras are global shutter, gen locked, strobe synced (with zero motion blur) and can record at 60fps.

We have the ability to scan seated subjects in a full 360-degree camera arrangement utilizing 60x machine vision cameras, as well as illuminating the subjects with 360-degree cross-polarized global illumination lighting array from 160x LED’s supplied by IDA-Tronic. The camera data is streamed RAW directly to NVME bifurcated storage arrays at very high speeds.

We use microsecond timing on our LED system to strobe sync the camera shutters to acquire perfect synchronization. This reduces motion blur as well as creates a comfortable acting environment for the talent.

Custom Photometric and Motion Scanning Reconstruction Pipeline

IR’s powerful automated reconstruction system was designed using a custom CLI backend working in unison with Epic’s awesome 3D reconstruction software called Reality Capture. The output from S0L’s motion scanned frame reconstructions feature super sharp clean mesh geometry. The system also captures detailed 360-degree diffusely lit (shadow free) UV skin texture maps, captured per frame. Featuring skin wrinkling and blood flow.

High Fidelity

Our scanning system does not miss a (heart!) beat. We can capture extremely nuanced and subtle emotive performances. The S0L system offers you the ability to bring your digital humans to life with a high level of accuracy and realism. These extensive data sets can translate 1:1 into the digital world. We strive to capture and provide high quality motion scan data.

With the unprecedented high-fidelity of our scanning system, your studio can replicate extremely life like digital humans with ease.

Training Data Sets for Machine Learning Applications

We’ve worked closely with machine learning companies over the last 5 years to develop a rigorous scanning pipeline to capture huge ranges of human emotion, including facial expressions, calibrated motions as well as dedicated performances. We based our capture pipeline on the research by MIT, HARVARD, META and Hjortsjö, Ekman and Friesen. We have the ability to help you build large data sets for machine learning applications.

Our system has the ability to capture up to 40-minutes of motion scan footage a day, uninterrupted. The S0L system is equipped with over 120 terabyte’s of direct storage.

For a typical “machine learning” project we would generally capture 20 minutes of footage which equals over 70 terabyte’s of RAW data. Which translates to 72,000 training shapes (scans), or roughly 4,320,000 captured images!

Our in-house storage facility has over 3 Petabytes of storage capacity. We’re also equipped with a super fast optical fibre line to be able to deliver high quality data to clients worldwide at blazingly fast speeds.

Extensive Motion Scan Data Sets

We are the driving force behind the scenes. We can do the heavy lifting, so that you don’t have to.

To be able to produce “hero quality” and sophisticated Face Training solutions as well as automated facial animation pipelines, it requires vast amounts of training data, translated from our motion scans.

Our systems are designed to facilitate this need, today.

There’s so much nuanced performance hidden in a RAW motion scan capture. From muscle twitching, to tendons activating, to skin sliding. Our systems can capture it all. We can do it reliably, on-time and without fail. So you don’t have to reinvent the wheel.

We can tailor our motion scanning sessions to cater to your pipeline. The possibilities are endless.

Let us know what your machine learning systems require!