Grayscale faults

  • The output by the code device causes distorted grayscale faults

  • The initial node does not have this fault situation.

  • I’m not sure if it’s a floating-point accuracy problem, but the current code device doesn’t support outputting double-floating-point precision graphs.

1 Like

Are you outputting an rgb image from your code device?

RGB data is stored as fp16. So it is reduced in precision compared to heightfield data. This is a great tradeoff for operating on actual image data, but it is not quite enough precision to use for geometry purposes.

The recommended workaround for the above right now is to instead export two heightfields from your code device.

2 Likes

It is true that FP16 is insufficiently accurate and has been solved

1 Like

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.