search

jmccormac / pySceneNetRGBD

Scripts showing how to work with the SceneNetRGBD dataset
GNU General Public License v3.0
178 stars 46 forks source link

depth definition #17

Closed versatran01 closed 6 years ago

versatran01 commented 6 years ago

I'm a bit confused about how to interpret depth from the dataset. The paper says

Depth is defined as the ray length from the camera origin to the first surface it intersects with

From my understanding, this is different from the depth map. Please advise.

jmccormac commented 6 years ago

Yes the use of euclidean depth here is different from the depth image produced from say a Kinect RGB-D camera. The depth values in a kinect measures the z-axis component of the point in the camera coordinate frame. A diagram showing that is available here

Basically if the camera views a flat wall head-on, in this dataset the corners of the image will be at a greater euclidean distance (the hypotenuse) than the center, but in a kinect camera they will all have the same depth values which is the distance the wall is along the z-axis (the adjacent).