Is there a way to improve hand detection accuracy?

The Hand Tracking SDK uses a deep learning model to detect hands and fingers, and detection accuracy depends mainly on the training images. The HTC development team is continuously collecting data and improving accuracy.

Hand detection, especially finger tracking, may vary based on your lighting condition and background. Here are some tips that may help improve accuracy:

Roll your sleeves and make sure your wrist is fully visible. This can highly improve finger detection accuracy.
On VIVE Pro and VIVE Cosmos, lift your hand a little higher since the vertical FOV (Field of View) of their front camera(s) is quite small.
Avoid places where the color of your background looks similar as your hand complexion. Choose a place where your hands can be detected against a high-contrast background.
The same goes for your lighting. Try to experiment with various lighting colors,. Make sure your room lighting is neither too dark (which will make it hard to distinguish your hand from your environment) nor too light (which may result in losing hand texture details).
If you're using an Android phone, make sure to turn your phone counterclockwise and hold it in landscape orientation.

Was this helpful? Yes No

Submit

Thank you! Your feedback helps others to see the most helpful information.