Parallelized Hausdorff-Based Matching

APMA2822B - Introduction to Parallel Computing on Heterogeneous(CPU+GPU) Systems

Final Project (May, 2019)

Yang Jiao [email protected]

Geng Yang [email protected]

Software Architecture

Dependencies

• OpenCV 3.4.1 (only for loading images)
• OpenMP 4.0
• CUDA 10.0.130

Build System

• Redhat 7 (Brown's CCV)

Instructions for compiling and running

This software can be built on Linux, macOS and Windows, the instructions are as follows.

1. Clone the repository.

$ [email protected]:Genggg/ParaHausdorff.git

2. If build on CCV, please load the following modules.

$ moudle load opencv/3.4.1
$ moudle load cuda/10.0.130

3. Generate MakeFile and Compile.

$ mkdir build
$ cd build
$ cmake ..
$ make

4. Modify the execution command in cuda.sh, in which the following args are the paths of target image and template respectively.

./main ../img/coins.jpg ../img/coin_T.png

5. Submit the task and check the result

$ sbatch cuda.sh
$ cat run.out