GPGPU - Simple Object Detection

Authors

• Moustapha Diop
• Mathieu Rivier
• Othman Elbaz
• Lucas Pinot

Running the project:

Running a video (in images format):

To run a a video file (that is given in a folder) as images use:

• your images have to be named thumb{num}.jpg (or change the name in the command bellow)

make build
make gpu_run path=../images/{your folder}/thumb*.jpg
make clean # to clean and delete the build folder

Running the program once (with our test images):

make build
make gpu_test
make clean # to clean and delete the build folder

Running the program once (with our test images) (CPU):

make build
make cpu_test
make clean # to clean and delete the build folder

Running the benchmarks (CPU):

make build
make bench_cpu
make clean # to clean and delete the build folder

Running the benchmarks (GPU):

make build
make bench_gpu
make clean # to clean and delete the build folder

Project Architecture

Basics:

The code is located in the src folder while the header files are in the includes folder.

The paper.pdf is the report that represents the final output and results of this project.

.
├── src
├── includes
├── images
├── visu
├── Makefile
└── paper.pdf

Code Architecture

src
├── CMakeLists.txt
├── bench
│   ├── bench_cpu.cpp
│   └── bench_gpu.cpp
├── cpu
│   ├── CMakeLists.txt
│   ├── bbox_cpu.cpp
│   ├── blur_cpu.cpp
│   ├── detect_obj_cpu.cpp
│   ├── main_cpu.cpp
│   ├── opening.cpp
│   └── threshold
│       ├── connexe_components_cpu.cpp
│       ├── otsu_cpu.cpp
│       └── threshold_cpu.cpp
├── gpu
│   ├── CMakeLists.txt
│   ├── bbox_gpu.cu
│   ├── blurr_gpu.cu
│   ├── detect_obj_gpu.cu
│   ├── difference_gpu.cu
│   ├── gray_scale_gpu.cu
│   ├── helpers_gpu.cu
│   ├── main_gpu.cpp
│   ├── opening_closing_gpu.cu
│   ├── threshold
│   │   ├── connexe_components_gpu.cu
│   │   ├── otsu_gpu.cu
│   │   └── threshold_gpu.cu
│   └── utils_gpu.cu
├── helpers_images.cpp
└── struct_utils.cpp

Different Outputs of the Program: