rodrigob/barinova_pedestrians_detection

Linux port of original Windows code from Olga Barinova

barinova_pedestrians_detection's Issues

Forest file doesn't exist

I run the command the do the testing:

./object_detection ~/Develop/barinova_pedestrians_detection-master/example/campus-config

It gives me an error:

Loading config file
Done
Reading images from host_folder == '/home/zhuofu/Develop/barinova_pedestrians_detection-master/example/tud-campus-sequence/'
Created output folder /home/zhuofu/Develop/barinova_pedestrians_detection-master/example/results/
Reading forest from file ...
Forest file doesn't exist!

My campus_config file:

// folder with input images
/home/zhuofu/Develop/barinova_pedestrians_detection-master/example/tud-campus-sequence/
\\
// output folder
/home/zhuofu/Develop/barinova_pedestrians_detection-master/example/results/
\\
// width of a bounding box at each scale corresponding to detection
40
\\
// height of a bounding box at each scale corresponding to detection
84
\\
// number of scales for multi-scale detection
5
\\
// resize images before detection by this coefficient
0.55
\\
//forest name
/home/zhuofu/Develop/barinova_pedestrians_detection-master/example/barpedestrian.dat
\\
//size of patch in a forest
16
\\
// half of maximum width of object in the data used for training the forest
30
\\
// half of maximum height of object in the data used for training the forest
45
\\
// bias of background cost, parameter of detection algorithm
8.5
\\
// penalty for adding a hypothesis, parameter of detection algorithm
1500
\\
//only patches with probabilities to belong to background lower than this threshold can vote
0.7
\\
//minimal allowed probability of a patch to belong to object, if probability is lesser the vote is ignored
0.00005
\\
// radius of blur for hough images
5
\\
// size of images at subsequent scales differ by coefficient 
0.85
\\
// maximum number of objects in an image
8

Missing boost dependencies

Even tough I installed all the boost libraries by calling
apt-get install libboost-all-dev
Still I get following error at compilation

/usr/bin/ld: cannot find -lboost_filesystem-mt
/usr/bin/ld: cannot find -lboost_system-mt
collect2: error: ld returned 1 exit status
make[2]: *** [object_detection] Error 1
make[1]: *** [CMakeFiles/object_detection.dir/all] Error 2
make: *** [all] Error 2

Indicated host_folder does not exists

When testing, running the command

 ./object_detection '/home/rnt/pedestrian/untouched/barinova_pedestrians_detection-master/example/campus-config' <b>

gives me output

Loading config file
Done
'eading images from host_folder == '/home/rnt/pedestrian/tud-campus-sequence/
IMPORTANT: Is your configuration file in Unix line ending format ? (Windows style line ending not supported)
A std::exception was raised: Indicated host_folder does not exists
terminate called after throwing an instance of 'std::invalid_argument'
  what():  Indicated host_folder does not exists
Aborted (core dumped)

My campus-config file is

// folder with input images
/home/rnt/pedestrian/tud-campus-sequence/
\\
// output folder
/home/rnt/results/
\\
// width of a bounding box at each scale corresponding to detection
40
\\
// height of a bounding box at each scale corresponding to detection
84
\\
// number of scales for multi-scale detection
5
\\
// resize images before detection by this coefficient
0.55
\\
//forest name
/home/rnt/pedestrian/untouched/barinova_pedestrians_detection-master/example/pedestrian.dat
\\
//size of patch in a forest
16
\\
// half of maximum width of object in the data used for training the forest
30
\\
// half of maximum height of object in the data used for training the forest
45
\\
// bias of background cost, parameter of detection algorithm
8.5
\\
// penalty for adding a hypothesis, parameter of detection algorithm
1500
\\
//only patches with probabilities to belong to background lower than this threshold can vote
0.7
\\
//minimal allowed probability of a patch to belong to object, if probability is lesser the vote is ignored
0.00005
\\
// radius of blur for hough images
5
\\
// size of images at subsequent scales differ by coefficient 
0.85
\\
// maximum number of objects in an image
8

Build errors with recent version of CMake (2.8.7)

I tried building the library in Ubuntu 12.04, OpenCV 2.2 (also attempted other configurations), and kept getting an error like the following:

/usr/bin/c++    -Wall -DNDEBUG -O3 -fopenmp -funroll-loops -march=native -mtune=native --fast-math -mfpmath=sse -msse3     CMakeFiles
/object_detection.dir/home/kit/Downloads/barinova_pedestrians_detection/ObjectDetection/main_run_greedy_pedestrians.cpp.o CMakeFiles/
object_detection.dir/home/kit/Downloads/barinova_pedestrians_detection/ObjectDetection/DenseGreedyDetection/GreedyDetection.cpp.o CMa
keFiles/object_detection.dir/home/kit/Downloads/barinova_pedestrians_detection/ObjectDetection/Forests/HoughForest.cpp.o CMakeFiles/o
bject_detection.dir/home/kit/Downloads/barinova_pedestrians_detection/ObjectDetection/Forests/IHoughForest.cpp.o CMakeFiles/object_de
tection.dir/home/kit/Downloads/barinova_pedestrians_detection/ObjectDetection/Forests/IHoughTree.cpp.o CMakeFiles/object_detection.di
r/home/kit/Downloads/barinova_pedestrians_detection/ObjectDetection/Forests/HoughTree.cpp.o CMakeFiles/object_detection.dir/home/kit/
Downloads/barinova_pedestrians_detection/ObjectDetection/Forests/Vote.cpp.o CMakeFiles/object_detection.dir/home/kit/Downloads/barino
va_pedestrians_detection/ObjectDetection/ImageWrapper/HoG.cpp.o CMakeFiles/object_detection.dir/home/kit/Downloads/barinova_pedestria
ns_detection/ObjectDetection/ImageWrapper/MultiImageTools.cpp.o CMakeFiles/object_detection.dir/home/kit/Downloads/barinova_pedestria
ns_detection/ObjectDetection/ImageWrapper/CRPatch.cpp.o  -o object_detection -rdynamic -lboost_filesystem-mt -lboost_system-mt -lgomp
 -ltbb -lrt -lpthread -lm -ldl
CMakeFiles/object_detection.dir/home/kit/Downloads/barinova_pedestrians_detection/ObjectDetection/main_run_greedy_pedestrians.cpp.o:
In function `HoG::~HoG()':
main_run_greedy_pedestrians.cpp:(.text._ZN3HoGD2Ev[_ZN3HoGD5Ev]+0x9): undefined reference to `cvReleaseMat'
main_run_greedy_pedestrians.cpp:(.text._ZN3HoGD2Ev[_ZN3HoGD5Ev]+0x12): undefined reference to `cvReleaseMat'
main_run_greedy_pedestrians.cpp:(.text._ZN3HoGD2Ev[_ZN3HoGD5Ev]+0x1b): undefined reference to `cvReleaseMat'
main_run_greedy_pedestrians.cpp:(.text._ZN3HoGD2Ev[_ZN3HoGD5Ev]+0x24): undefined reference to `cvReleaseMat'
main_run_greedy_pedestrians.cpp:(.text._ZN3HoGD2Ev[_ZN3HoGD5Ev]+0x2d): undefined reference to `cvReleaseMat'

etc...

The issue was cmake not finding the opencv libraries properly.

I needed to change pkg_check_modules(opencv REQUIRED opencv>=2.1) to find_package( OpenCV 2.1 REQUIRED )

and target_link_libraries(... ${opencv_LIBRARIES} ) to target_link_libraries(... ${opencv_LIBS} )

Can not download the tud-campus and tud-crossing test sequences

Download the tud-campus and tud-crossing test sequences, every time I try to get into the website, I saw 'EOF'. Where can I get the data?

Are there any YouTube videos or pictures of the project in practice?