This is a fork of the BING Objectness code from www.mmcheng.net/bing for Linux and OSX.
It should also compile on Windows as intended by the original author (not tested).
Note: OpenMP is not yet supported by the default LLVM compiler in OSX, hence it is disabled for OSX.
Nevertheless, the code can still be compiled and executed successfully in OSX.
It implements the techniques described in the following paper:
BING: Binarized Normed Gradients for Objectness Estimation at 300fps.
Ming-Ming Cheng, Ziming Zhang, Wen-Yan Lin, Philip Torr.
IEEE CVPR, 2014. pdf
These are the results that I obtained using the fast detection variant
1:0.25,0.304 10:0.481,0.449 100:0.794,0.607 1000:0.958,0.669
2000:0.969,0.673 3000:0.969,0.673 4000:0.969,0.673 5000:0.969,0.673
Download the VOC 2007 train data and put them in the VOC2007 directory.
Do not copy the ImageSets directory from the VOC dataset, retain the version supplied with this code.
Download the [VOC2007 test data] (http://pascallin.ecs.soton.ac.uk/challenges/VOC/voc2007/VOCtest_06-Nov-2007.tar)
Move all images from /JPEGImages folder to the earlier extracted VOC2007/JPEGImages folder.
Unzip the Annotations.zip file in the VOC2007 folder.
-Or-
Download an OpenCV readable version of VOC 2007 annotations.
Set OPENCV_DIR set as an environment variable in the shell if it is not present in the standard directories.
mkdir build
cd build
cmake ..
make