Code for combining Partial Object View with a CNN generated Completion
Eigen: a recent install is required, default from apt-get for 14.04 is not new enough
CUDA: Optional, but speeds up reconstruction time.
Usage: mesh_reconstruction <binvox file> <pcd file> <output file> <options>
Allowed options:
--help produce help message
--feature-detection Toggle feature handling
--cuda Toggle CUDA option
--feature-threshold arg Increasing raises feature sensitivity. Default: 0.75
--corner-threshold arg Decreasing raises feature sensitivity. Default: 0.8
<binvox file>
is the output of running our CNN, normally a 40x40x40 occupancy map representing our hypothesis about which voxels are occupied by the object.
<pcd file> is the captured partial view of the target object, this only contains the points visible to the camera and tends to be higher resolution than the completion we are combining it with
where to place to merged partial view and completion