For the original description, see the original repo.

This repo contains the mex files needed for running Ma's code on a Linux system, as well as the changes made to the files that produced those mex files. Tested on Ubuntu 20.04 with g++/gcc 9.4.0, MATLAB R2022a.

Steps performed to accomplish this (generally, these follow the instructions from the "Troubleshooting" section of the README of this repo; below, the sr-metric dir in paths can be replaced with sr-metric-fixed-for-linux, if you carry out the below instructions using this repo):

1. Run sr-metric/external/matlabPyrTools/MEX/compilePyrTools.m.
2. After execution, compilePyrTools.m generated new versions of 4 mex files: upConv.mexa64, pointOp.mexa64, range2.mexa64, histo.mexa64. Copy these to the parent dir (i.e., to sr-metric/external/matlabPyrTools/). These files already exist in that parent dir, so you'll need to overwrite those.
3. Change line 82 in the file sr-metric/external/randomforest-matlab/RF_Reg_C/src/mex_regressionRF_predict.cpp to plhs[0]=mxCreateNumericMatrix(n_size,1,mxDOUBLE_CLASS,mxREAL);.
4. Change the sr-metric/external/randomforest-matlab/RF_Reg_C/compile_linux.m so that it looks like so. Note that:
   • The -o switch won't work, so use -output instead.
   • You must give the correct paths to the C++ files, i.e., precede them with src/, provided you will be running compile_linux.m from its directory, that is, from sr-metric/external/randomforest-matlab/RF_Reg_C/.
   • You might need to supply the full path to mex for your MATLAB installation if make mex won't work.
5. Run compile_linux.m from its directory. This will generate the file mexRF_predict.mexa64, which is the file required to resolve the undefined function or variable mexRF_predict error.
6. That is it, the mexRF_predict.mexa64 file should stay in sr-metric/external/randomforest-matlab/RF_Reg_C and doesn't need to be copied anywhere. You can now execute the code for Ma's metric, e.g., run demo.m.