This repository contains
- the code to compile the subsampling executables that are used by benchmark users to create submission files.
- the code to generate the ground truth files as needed by the benchmark website for evaluation
- clone HighFive (tested with commit 5d88d87155a14cd694ee342c8f77e61815d2810a)
- create build folder:
mkdir build && cd build
- execute cmake:
cmake ..
- execute make:
make -j <numprocesses>
The code was tested with gcc 11.3.0, cmake 3.22.1 on Ubuntu 22.
The build result are three executables; disp1_subsampling, disp2_subsampling and flow_subsampling, which are used to generate the corresponding submission files, e.g. via ./disp1_subsampling <path to disp1 results>
.
Please note that the true subsampling random numbers are withheld and have to be added in the *.cpp files in order to create correct submission executables.
Additionally, this repository contains the code to generate subsampled ground truth files (eval files) that are used by the benchmark website evaluation code:
- Install the required python libraries numpy and h5py.
- Set the environment variable SPRING_TESTPATH to the test split directory of the spring dataset that contains the ground truth files
- Run
python create_evalfiles.py
- The resulting eval_*.hdf5 files are needed by the benchmark evaluation code.