chenjiechun / Branch-structured-Face-detector-SDK

Branch-structured face detector: an efficient detector that can obtain higher performance but use less memory.

Branch-structured-Face-detector-SDK

Brief description

This is a detector that can make a trade-off between efficiency and performance for face detection. It was developed on the basis of our paper: Chen, Jc., Wang, J., Zhao, Lp. et al. Branch-structured detector for fast face detection using asymmetric LBP features. SIViP 14, 1699–1706 (2020). https://doi.org/10.1007/s11760-020-01710-7 . The main contributions of our manuscript are as follows:

• We designed a branch-structured detector.
• We proposed a pixel-level image feature named Asymmetric LBP feature (ALBP).
• We proposed a method for introducing ALBP features into a branch-structured face detector.

Performance and Efficiency Evaluation

Fig 1. Discrete ROC curve for our detector on FDDB Dataset.

Table 1. Detection speed of our detector

CPU	Image Size	Step	Scale factor	Needed memory	Threads	Min Face	FPS
i7-7700	640*480	3	1.2	15MB	4	40*40	316(a.jpg), 197(b.jpg)
i7-7700	640*480	3	1.2	15MB	4	80*80	784(a.jpg), 572(b.jpg)
i7-7700	640*480	3	1.2	15MB	2	40*40	301(a.jpg), 186(b.jpg)
i7-7700	640*480	3	1.2	15MB	2	80*80	767(a.jpg), 554(b.jpg)
i7-7700	640*480	3	1.2	15MB	1	40*40	261(a.jpg), 167(b.jpg)
i7-7700	640*480	3	1.2	15MB	1	80*80	741(a.jpg), 529(b.jpg)

a.jpg	b.jpg

Requirement

• Windows (Windows7 has been validated).
• C++ compiler [Visual studio 2013 (Community version) has been validated].

How to Create an face detection program with Visual Studio 2013

1. Create a project: FILE -> New -> Project... -> Visual C++ -> Win32 -> Win32 Console Application.

2. Include necessary header files in .cpp files.

   For instance, we create a project named "test". Then, a .cpp file named "test.cpp" will be created automatically. In the test.cpp file, header files should be included as follows:

   #include "opencv2/highgui/highgui.hpp"
   #include "opencv2/imgproc/imgproc.hpp"
   #include "bsFaceDetector.h"

3. Add "<SDK_Folder>\include" to Additional Include Directories: (Project) Properities -> Configuration Properties -> C/C++ -> General -> Additional Include Directories.

   <SDK_Folder> represents the folder which stores SDK files. For instance, the SDK files are stored at d:\BSFD_SDK, <SDK_Folder> would represent "d:\BSFD_SDK" .

4. Add "<SDK_Folder>\lib\x64" (or "<SDK_Folder>\lib\x86") to Additional Library Directories: (Project) Properities -> Configuration Properties -> Linker -> General -> Additional Libray Directories.

5. Add necessary Static Libraries to Additional Dependencies: (Project) Properities -> Configuration Properties -> Linker -> Input -> Additional Dependencies. The necessary Static Libraries are:

   • opencv_highgui2411.lib
   • opencv_core2411.lib
   • bsFaceDetector.lib
   • seeta_facedet_lib.lib

6. Build project.

7. Copy model files to the folder which stores the executable file (.exe) created by visual studio 2013. The model files that should be copied are:

   • cascadeModels.bin
   • seeta_fd_frontal_v1.0.bin

8. Copy the .dll files that are stored in "<SDK_Folder>\Dll\x64" (or "<SDK_Folder>\Dll\x86") to the folder which stores the executable file (.exe) created by visual studio 2013.

9. Run the executable file from Command Prompt.

Contact with us

Email: [e-mail]
QQ: 418044891