Facebook, the mogul of online socializing, tried to bring social experiences to virtual reality by launching Facebook Spaces a year back. Facebook Spaces is an online gathering of VR enthusiasts wherein one can engage with their friends and family in the form of customizable cartoon-like avatar characters.
In order to enhance the user’s social VR experience further, Facebook-owned Oculus is working to improve hand tracking technology. The hope is to allow for more accurate avatars and intuitive controls in virtual reality.
Maria Fernandez Guajardo, head of product management on core tech, recently offered a glimpse into the ongoing research at Oculus’s R&D Division.
During the presentation, Guajardo unveiled a hand-tracking system (based on computer-vision) that incorporates a host of smart technologies such as machine-learning algorithms. She said that this hand-tracking system is “far more accurate than any method before for tracking a single hand, two hands, and hand-object interactions.”
In order to operate with such accuracy, Oculus makes use of a marker-based system that records hand interactions and movements in higher detail. The recorded data is then condensed into 2D imagery, thereby setting up a convolutional neural network. This neural network identifies the positions of markers across a large set of hand imagery which will help the tracking system figure out how the hand should look under different marker position scenarios.
Apparently, this trained system can then be fed to the markerless camera input of virtual reality headsets and allow them to track hand-movements based on the information the VR headset cameras have captured.
Results of Oculus’ Hand-Tracking Solutions
Having tested this system, Oculus claims to have an accuracy rating much higher than its competitors. In the case of single-hand tracking, the company claims to have attained an exemplary success rate of 100% against the 90.49% rate achieved through other hand-tracking methods.
Moreover, the above chart shows that there is a significant jump in the results and accuracy of two-handed tracking and hand-to-object tracking interactions. It seems that as these solutions come to the market in the near future, competitors in this hand-tracking segment, such as Leap Motion, will certainly have a run-for-their-money.
The hand-tracking solution, when implemented, will certainly add a new sense of immersion for touch and feel in the VR Porn games.