Issue with getting results about vacancy

[unclearness]

The simplest way is passing the silhouette image in black and white as in examples.cc

Just in case, the image you posted below in red and blue is NOT SDF (32-bit float one channel in this implementation) itself; It's a visualization of SDF in 8-bit uchar (0-255) 3 channels.
It shouldn't be an input to this code.

Reasons for failures with valid image inputs would be either your camera parameters or voxel initialization range (set in bb_max and bb_min) is wrong.
Note that this code uses the OpenCV camera coordinate. If you want to pass the ones from CG software such as Blender to this code, you must convert them correctly.