search

alexsax / 2D-3D-Semantics

The data skeleton from Joint 2D-3D-Semantic Data for Indoor Scene Understanding
http://3dsemantics.stanford.edu
Apache License 2.0
464 stars 67 forks source link

Depths images look black #8

Closed longlouisly closed 7 years ago

longlouisly commented 7 years ago

Thank you for providing such a great dataset. Question: In your sample figure, the depth map looks smooth with many grayscale variations. However, when I view the data, most are black/white, with very little grayscale or smoothness. Is this intended or am I reading the pngs incorrectly?

alexsax commented 7 years ago

You're welcome! This is intended behavior! In the data, the images are 16-bit and so look black. They do contain the correct depth information, and for visualizing I recommend either using the log-depth or else rescaling the values linearly based on the max (nonmissing) depth in the image. In the figure you linked, we used the log(depth) just to make the viz easier.

longlouisly commented 7 years ago

Thanks for your quick response. I also realize now that the missing values were set to white (255).

alexsax commented 7 years ago

Yeahp, missing values are set to the maximum possible value. Also, make sure that you're using 16-bit channels (and that the max value is 65535) so that you get the full depth resolution!