libfacedetection

This is a binary library for face detection in images. The 32-bit and 64-bit dll files are provided.

examples/libfacedetect-example.cpp shows how to use the library.

Comparison on Windows

• 640x480 image size (VGA), scale=1.2, minimal window size = 48
• Intel(R) Core(TM) i7-4770 CPU @ 3.4GHz
• Multi-core parallelization is enabled for the four methods.

Comparison on iPhone

• 640x480 image size (VGA), scale=1.2, minimal window size = 48
• Multi-core parallelization is disabled.
• C programming language, and no SIMD instruction is used.

Comparison on ARM

• 640x480 image size (VGA), scale=1.2, minimal window size = 48
• NVIDIA TK1 "4-Plus-1" 2.32GHz ARM quad-core Cortex-A15 CPU
• Multi-core parallelization is disabled.
• C programming language, and no SIMD instruction is used.

The dll cannot run on ARM. The library should be recompiled from source code for ARM compatibility. If you need the source code, a commercial license is needed.

Evaluation

FDDB: http://vis-www.cs.umass.edu/fddb/index.html

• scale=1.08
• minimal window size = 16
• the heights of the face rectangles are scaled to 1.2 to fit the ground truth data in FDDB.

Tip

• Do NOT use the functions in multiple threads. The memory used in the functions is not protected according to multiple threads.

Author

• Shiqi Yu shiqi.yu@gmail.com Computer Vision Institute, Shenzhen University, China

Contributors

• Shengyin Wu, Shenzhen University, China
• Dong Xu, Shenzhen University, China