Around the holidays, Desi and Wanda shared a video on YouTube noting a performance issue with HP facial tracking software. Since then, the team at HP has been working hard to understand the issue and improve the software.
We now have an update available ( Here's the desktop version link, and here's the notebook version link). A special thanks to Desi and Wanda for helping us test it. Let me explain how it works.
Facial tracking software identifies a face in two ways. It measures the contrast between the eyes and upper cheeks and the eyes and the bridge of the nose. The image below illustrates how those patterns look in ideal lighting.
In environments with uneven light or bright background lighting (which creates foreground shadows), those patterns become less clear. Variance in the way light reflects off different skin tones can reduce the clarity of the patterns even more.
We have done a couple things with the update to help improve the user experience in these situations. The first is to automatically enable backlight compensation, a feature that digitally brightens foreground shadows. It’s worth noting that this is enabled on the video feed the algorithm analyses, not the one displayed on the screen. You may remember that we recommended manually enabling this feature when the video first surfaced.
The second is to slightly loosen the criteria the software uses to identify patterns. In a sense, this works similarly to the way the human eye can recognize a shape even when a small part of it might be missing.
We will continue to tune the algorithm to improve the webcam experience. If you still encounter issues, please let me know. You’re also welcome to contact us on Twitter, Facebook and YouTube. As always, your feedback is important to us, and we will continue to listen.
Around the holidays, Desi and Wanda shared a video on YouTube noting a performance issue with HP facial tracking software [INSERT LINK TO PREVIOUS POST]. Since then, the team at HP has been working hard to understand the issue and improve the software.
We now have an update available (desktop version, notebook version). A special thanks to Desi and Wanda for helping us test it. Let me explain how it works.
Facial tracking software identifies a face in two ways. It measures the contrast between the eyes and upper cheeks and the eyes and the bridge of the nose. The image below illustrates how those patterns look in ideal lighting.
In environments with uneven light or bright background lighting (which creates foreground shadows), those patterns become less clear. Variance in the way light reflects off different skin tones can reduce the clarity of the patterns even more.
We have done a couple things with the update to help improve the user experience in these situations. The first is to automatically enable backlight compensation, a feature that digitally brightens foreground shadows. It’s worth noting that this is enabled on the video feed the algorithm analyses, not the one displayed on the screen. You may remember that we recommended manually enabling this feature when the video first surfaced.
The second is to slightly loosen the criteria the software uses to identify patterns. In a sense, this works similarly to the way the human eye can recognize a shape even when a small part of it might be missing.
We will continue to tune the algorithm to improve the webcam experience. If you still encounter issues, please let me know. You’re also welcome to contact us on Twitter, Facebook and YouTube. As always, your feedback is important to us, and we will continue to listen.
