Skip to main content

AUTO-MNET (Body Part Tracking)

A 3D human pose estimator for edge devices. Automotive Challenge - Body Part Tracking


Business Category

What is the challenge that is being addressed?

We have developed a method for 3D realtime human motion capture from monocular visual 2D data. We foresee that vision-enabled cars that are able to capture humans in 3D will be able to understand and anticipate the state and needs of drivers thus reducing the chances of motor accidents that are among the leading causes of death globally. We envision our technology having the impact of a modern day seat-belt or airbag that will non-intrusively and transparently act first as a passive safety mechanism while also leading the way to a futuristic driving experience facilitating effective human-machine collaboration.

What is the AI solution the project plans to implement?

Our method estimates the articulated 3D pose of human bodies from monocular RGB sources in real-time using MocapNETs. The method does not just regress 3D coordinates for recovered joints, but instead the full kinematic solution of the human body. Due to its unique design, compositionality and compact Bio Vision Hierarchy (BVH) output it can not only be used in automotive applications but also a wide variety of other applications.  

How will BonsAPPs support you in implementing this solution?

We believe that a European AI Marketplace will have a very positive effect for all involved parties and for society as a whole since matching producers to consumers is an essential function of all economic activity.  The tools offered can help porting solution across different heterogeneous platforms. As a result BonsAPPs has the potential to disrupt the current global AI market by creating a direct line of exchange between AI talents and SMEs.