xiaoerlaigeid / libfacedetection

A fast binary library for face detection and face landmark detection in images. The face detection speed can reach 1500FPS. You can use it free of charge with any purpose.

libfacedetection

This is a binary library for face detection and face landmark detection in images. The 32-bit and 64-bit dll files are provided. To achieve better performance, the 64-bit dll is recommended.

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

Comparison on Windows

Method	Time	FPS	Time	FPS	Time	FPS	Misc
	Win32	Win32	X64	X64	X64	X64
	Single-thread	Single-thread	Single-thread	Single-thread	Multi-thread	Multi-thread
OpenCV	--	--	--	--	12.33ms	81.1	Yaw angle: -60 to 60 degrees
frontal	2.92ms	342.5	2.41ms	414.9	0.652ms	1533.1	Yaw angle: -60 to 60 degrees
frontal-surveillance	3.83ms	261.1	3.37ms	269.7	0.944ms	1059.8	Yaw angle: -70 to 70 degrees
multiview	7.12ms	140.4	5.81ms	172.1	1.597ms	626.4	Yaw angle: -90 to 90 degrees
multiview_reinforce	10.95ms	91.3	9.15ms	109.3	2.725ms	367.0	Yaw angle: -90 to 90 degrees

• Face detection only, and no landmark detection included.
• 640x480 image size (VGA), scale=1.2, minimal window size = 48.
• Intel(R) Core(TM) i7-4770 CPU @ 3.4GHz.
• OpenCV classifier file: haarcascade_frontalface_alt.xml

Comparison on ARM

Method	Time	FPS	Misc
frontal	12.5ms	80.0	Yaw angle: -60 to 60 degrees
frontal-surveillance	15.7ms	63.7	Yaw angle: -70 to 70 degrees
multiview	27.8ms	36.0	Yaw angle: -90 to 90 degrees
multiview_reinforce	38.4ms	26.0	Yaw angle: -90 to 90 degrees

• Face detection only, and no landmark detection included.
• 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.

Author

• Shiqi Yu, [email protected]

Contributors

• Jia Wu
• Shengyin Wu
• Dong Xu

---

• The result image was taken by Chloé Calmon.