[question] (1) batch depth images and (2) coordinates in image

[ywpkwon]

Thanks for this great tool! Please let me ask two questions.

(1) I tried to extract screenshots of all car models successfully with save model screenshots all 3dw car. How can I save the depth images of all car models? If so, how can I save those in a batch mode?

(2) The output csv file informs bbox and camera position/target/up vectors. Let's say a corner point of a bounding box is (a, b, c). Is there a way to calculate its pixel coordinate in a screenshot image?

[msavva]

Apologies for the long delay in responding.

1. We currently do not have a feature for depth image rendering, though that can be implemented by adjusting the rendering code, which is based on JMonkeyEngine3. The relevant part of the code is https://github.com/ShapeNet/shapenet-viewer/blob/master/src/main/scala/edu/stanford/graphics/shapenet/jme3/viewer/ScreenshotAppState.java#L174 where the screenshot is saved out, and you would probably want to extract the depth buffer and save that into a screenshot in addition to the. It is probably useful to refer to the documentation of the JMonkeyEngine API at http://javadoc.jmonkeyengine.org/ in order to find the details of how to do this.

2. You could compute pixel coordinates by saving the camera projection matrix and applying it to the 3D corner point position to get the corresponding image pixel coordinates. Saving of the camera projection matrix would again have to be implemented since it is not something that we currently provide. The relevant code location would be again, the one above.