libfacedetection-vs2015

This is ready to use vs2015 version of the original libfacedetect - https://github.com/ShiqiYu/libfacedetection.
You can use this libfacedetection-vs2015 to detect face in image/images/webcam and also for benchmarking face detection performance in AFW,PASCAL,FDDB,UFDD and WIDER face dataset.

1. Test webcam

$libfacedetection.exe -mode=0 -webcam=0

2. Test single image

$libfacedetection.exe -mode=1 -path=../image/1.jpg

3. Test image lists

$libfacedetection.exe -mode=2 -path=../image/

4. Evaluation in benchmark dataset, detection files will be stored in "detections" folder.

a) afw dataset
$libfacedetection.exe -mode=3 -dataset=AFW -path=/path/to/afw/dataset/

b) PASCAL dataset
$libfacedetection.exe -mode=3 -dataset=PASCAL -path=/path/to/pascal/dataset/

c) FDDB dataset
$libfacedetection.exe -mode=3 -dataset=FDDB -path=/path/to/fddb/dataset/

d) WIDER_val dataset
#libfacedetection.exe -mode=3 -dataset=WIDER_VAL -path=/path/to/wider/validation/dataset/

e) UFDD dataset
#libfacedetection.exe -mode=3 -dataset=UFDD -path=/path/to/UFDD/validation/dataset/

CNN-based Face Detection on Windows

Method	Time	FPS	Time	FPS
	X64	X64	X64	X64
	Single-thread	Single-thread	Multi-thread	Multi-thread
cnn (CPU, 640x480)	71.44ms	14.0	16.91ms	59.1
cnn (CPU, 320x240)	18.21ms	54.9	4.60ms	217.4

Face detection only, and no landmark detection included.
Minimal face size ~10x10
Intel(R) Core(TM) i7-7700 CPU @ 3.6GHz.

CNN-based Face Detection on ARM Linux (Raspberry Pi 3 B+)

Method	Time	FPS	Time	FPS
	Single-thread	Single-thread	Multi-thread	Multi-thread
cnn (CPU, 640x480)	593.86ms	1.68	183.85ms	5.44
cnn (CPU, 320x240)	140.50ms	7.12	45.48ms	21.99
cnn (CPU, 160x120)	30.15ms	33.17	10.75ms	92.99
cnn (CPU, 128x96)	20.20ms	49.49	6.73ms	148.53

Face detection only, and no landmark detection included.
Minimal face size ~10x10
Raspberry Pi 3 B+, Broadcom BCM2837B0, Cortex-A53 (ARMv8) 64-bit SoC @ 1.4GHz

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.

Binary library for ARM

The binary evaluation library for ARM can be downloaded at https://github.com/OAID/YSQfastfd . The detection functions can only be called about 2000 times for evaluation.

Accuracy 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, shiqi.yu@gmail.com

Contributors

Jia Wu
Shengyin Wu
Dong Xu

The result image was taken by Chloé Calmon.

ghimiredhikura/libfacedetection-vs2015