Render depth from nerf perspective

[EsbernTK]

Hi Im using the pyngp package to render out different views of my nerf scene, however i also need the depth for each of those views, how would i go about doing that in the easiest way? I have looked at the render mesh method, but it doesnt seem quite what i need. I also saw that in the .cu files there were some render modes, but i cannot find the way to activate the depth render mode in the pyngp package.

[Vincentqyw]

setting render mode to depth:

testbed.render_mode = ngp.Depth

see more here: https://github.com/NVlabs/instant-ngp/discussions/600

[EsbernTK]

Thank you very much :D A follow up question; im using the google colab notebook that is linked to in this repository, to run the code, and i cannot find where the pyngp package files actually are compiled to in the linux environment. I know it is python 3.7 and i have looked in both /etc/python3.7, /usr/lib/python3.7 and /usr/local/lib/python3.7 but cannot find it. Knowing where the actual package is would allow me to just look up the capabilities of the library directly in the package files.

[saunair]

Hi @Vincentqyw after capturing the depth, I see that the values are between 0 and 255. How do I get the maximum depth that I can use to normalize the values?

[kafei123456]

Hi @Vincentqyw after capturing the depth, I see that the values are between 0 and 255. How do I get the maximum depth that I can use to normalize the values?

Excuse me. I m using the same way to get the depth image, but the image have 4 channels(may be RGB-D or RGBA ?). I wonder how you capture the 1 channels depth image.

[CaptainEven]

setting render mode to depth:

testbed.render_mode = ngp.Depth

see more here: #600

Hi, i've got a 4 channel image:W×H×4 depth image after set

testbed.render_mode = ngp.Depth

How to get a 16 bit png one channel depth map?

[CaptainEven]

Hi @Vincentqyw after capturing the depth, I see that the values are between 0 and 255. How do I get the maximum depth that I can use to normalize the values?

Excuse me. I m using the same way to get the depth image, but the image have 4 channels(may be RGB-D or RGBA ?). I wonder how you capture the 1 channels depth image.

I got the same question, have you figured it out yet?