LossCalculatorTSDF

Loss Calculator TSDF

The LossCalculator generates the weights used for the loss shaping as described in the paper.

Build

HDF5

As you have build SDFGen before, you can now reuse the HDF5 package. For that, you need to set the HDF5_DIR in the CMakeLists.txt file as before.

TCLAP

Update the path in the CMakeLists.txt, the same as in SDFGen.

Building of LossCalculator

After updating the paths, you only have to build the current project with the given CMakeLists.txt.

mkdir cmake
cd cmake
cmake -DCMAKE_BUILD_TYPE=RELEASE .. 
make -j 8

Usage

You can now use the LossCalculator, which takes three arguments, the first is a concatenation of path with a comma. The second contains the result in our example always 512 and in the last the maximum amount of threads.

./LossCalculator -p data/510bb6a0e4dbe783109adc01b05d8c32/voxelgrid/output_0.hdf5,data/510bb6a0e4dbe783109adc01b05d8c32/voxelgrid/output_1.hdf5 -r 512 -t 8 

Again, we provide here also a script to do this automatically for the generated output_?.hdf5 containers.

If you have used the shell script from above and generate the voxelgrids with the generate_tsdf_volumes.py. Then you can now use without changing anything else the script generate_loss_values.py.

Be aware, that the four threads already need around 15 GB memory. So only increase the amount of the threads if you have enough memory on your system.