gMF

CUDA based mean-field inference for dense CRF

Build:

mkdir build cd build cmake .. make -j12

Multi-class segmentation demo:

./gMF ../data/2007_000129.jpg ../data/2007_000129.png ../data/2007_000129_test.png ./gMF <input image> <input unary> <output image>

Play with the parameters in demo.cpp, have fun!

Binary segmentation demo:

./bSeg ../data/me.mp4 ../data/msk.ppm ./bSeg <input video> <mask image>

'x' to next frame 'z' to previous frame 'q' to exit

Binary segmentation demo, with grascale per-pixel likelihood input:

cd ../data/ sh batch_binary.sh

then go to the data/binary folder to see results

Name		Name	Last commit message	Last commit date
Latest commit History 29 Commits
ORUtils		ORUtils
data		data
gMF_Lib		gMF_Lib
gSLIC_Lib		gSLIC_Lib
CMakeLists.txt		CMakeLists.txt
NVTimer.h		NVTimer.h
README.md		README.md
batch.cpp		batch.cpp
demo.cpp		demo.cpp
demo_backup.cpp		demo_backup.cpp
demo_binary.cpp		demo_binary.cpp
demo_bseg.cpp		demo_bseg.cpp
demo_gslic.cpp		demo_gslic.cpp
image_helper.h		image_helper.h