depthai-experiments icon indicating copy to clipboard operation
depthai-experiments copied to clipboard

Published 20 hours ago • verified publisher icon luxonis

Getting float depth image

Open fred107 opened this issue 2 years ago • 1 comments

Hi! I am trying to process depth images with our openvino blob model with OAK-D. According to the openvino guide, it is said that VPU doesn't support UINT16 input. So, it looks like the floating point conversion for UINT16 depth image is necessary in my case. Can I do such a type casting in a depth AI pipeline and directly feed them into our openvino model?

fred107 avatar Apr 20 '22 07:04 fred107

Hello @fred107 , Great question! We are working on supporting FP16 depth output (option on StereoDepth node), but in the meanwhile you can read how you can "import" INT16 depth into the custom model - tutorial here. Thoughts? Thanks, Erik

Erol444 avatar Apr 20 '22 11:04 Erol444