A python script that converts a depth map to a 3D mesh in .obj format.
Visualizing depth maps as a 3D mesh rather than a 2D grayscale image gives a better clue of what goes wrong in depth estimation.
Most papers on depth estimation visualize their results as a 3D mesh. However, they don't provide any code to do so. For this reason, I implemented this mini-script.
- Python 3.2+
- OpenCV
The simplest usage is to put the script in the same folder as depth map and, optionally, the texture and then call:
python DepthToObj.py --depthPath [DEPTH_NAME_OR_PATH] --texturePath [TEXTURE_NAME_OR_PATH]
required:
--depthPath – path to the depth file. Depth file is required to be stored in a 16bit integer .png format in millimeter scale (kinect). It is then converted to float and divided by 1000 to obtain meters. Defaults to 'depth.png'.
optional:
--texturePath – path to the texture file. Defaults to empty string. If not defined, .mtl will not be created.
--objPath – output path and name of .obj file. Defaults to model.obj. If path or file doesn't exist, they will be created.
--mtlPath – output path and name of .mtl file. Defaults to model.mtl. If path or file doesn't exist, they will be created. If texturePath was not defined, this file will not be created.
--matName – name of material to be created with the texture. Defaults to colored.