An open source library for face detection in images. The face detection speed can reach 1000FPS.

Last update: Dec 27, 2022

Related tags

Overview

libfacedetection

This is an open source library for CNN-based face detection in images. The CNN model has been converted to static variables in C source files. The source code does not depend on any other libraries. What you need is just a C++ compiler. You can compile the source code under Windows, Linux, ARM and any platform with a C++ compiler.

SIMD instructions are used to speed up the detection. You can enable AVX2 if you use Intel CPU or NEON for ARM.

The model files are provided in src/facedetectcnn-data.cpp (C++ arrays) & the model (ONNX) from OpenCV Zoo. You can try our scripts (C++ & Python) in opencv_dnn/ with the ONNX model. View the network architecture here.

examples/detect-image.cpp and examples/detect-camera.cpp show how to use the library.

The library was trained by libfacedetection.train.

How to use the code

You can copy the files in directory src/ into your project, and compile them as the other files in your project. The source code is written in standard C/C++. It should be compiled at any platform which supports C/C++.

Some tips:

Please add facedetection_export.h file in the position where you copy your facedetectcnn.h files, add #define FACEDETECTION_EXPORT to facedetection_export.h file. See: issues #222
Please add -O3 to turn on optimizations when you compile the source code using g++.
Please choose 'Maximize Speed/-O2' when you compile the source code using Microsoft Visual Studio.
You can enable OpenMP to speedup. But the best solution is to call the detection function in different threads.

You can also compile the source code to a static or dynamic library, and then use it in your project.

How to compile

CNN-based Face Detection on Intel CPU

Method	Time	FPS	Time	FPS
	X64	X64	X64	X64
	Single-thread	Single-thread	Multi-thread	Multi-thread
cnn (CPU, 640x480)	58.06ms.	17.22	12.93ms	77.34
cnn (CPU, 320x240)	13.77ms	72.60	3.19ms	313.14
cnn (CPU, 160x120)	3.26ms	306.81	0.77ms	1293.99
cnn (CPU, 128x96)	1.41ms	711.69	0.49ms	2027.74

Minimal face size ~10x10
Intel(R) Core(TM) i7-1065G7 CPU @ 1.3GHz

CNN-based Face Detection on ARM Linux (Raspberry Pi 4 B)

Method	Time	FPS	Time	FPS
	Single-thread	Single-thread	Multi-thread	Multi-thread
cnn (CPU, 640x480)	492.99ms	2.03	149.66ms	6.68
cnn (CPU, 320x240)	116.43ms	8.59	34.19ms	29.25
cnn (CPU, 160x120)	27.91ms	35.83	8.43ms	118.64
cnn (CPU, 128x96)	17.94ms	55.74	5.24ms	190.82

Minimal face size ~10x10
Raspberry Pi 4 B, Broadcom BCM2835, Cortex-A72 (ARMv8) 64-bit SoC @ 1.5GHz

Performance on WIDER Face

Run on default settings: scales=[1.], confidence_threshold=0.3, floating point:

AP_easy=0.856, AP_medium=0.842, AP_hard=0.727

Author

Shiqi Yu, [email protected]

Contributors

All contributors who contribute at GitHub.com are listed here.

The contributors who were not listed at GitHub.com:

Jia Wu (吴佳)
Dong Xu (徐栋)
Shengyin Wu (伍圣寅)

Acknowledgment

The work was partly supported by the Science Foundation of Shenzhen (Grant No. 20170504160426188).

Citation

We published a paper on face detection to evaluate different methods. This project has also been evaluated in the paper.

@article{facedetect-yu,
 author={Yuantao Feng and Shiqi Yu and Hanyang Peng and Yan-ran Li and Jianguo Zhang}
 title={Detect Faces Efficiently: A Survey and Evaluations},
 journal={IEEE Transactions on Biometrics, Behavior, and Identity Science},
 year={2021}
 }

The paper can be open accessed at https://ieeexplore.ieee.org/document/9580485

The loss used in training is EIoU, a novel extended IoU. More details can be found in:

@article{eiou,
 author={Peng, Hanyang and Yu, Shiqi},
 journal={IEEE Transactions on Image Processing}, 
 title={A Systematic IoU-Related Method: Beyond Simplified Regression for Better Localization}, 
 year={2021},
 volume={30},
 pages={5032-5044},
 doi={10.1109/TIP.2021.3077144}
 }

The paper can be open accessed at https://ieeexplore.ieee.org/document/9429909.

Comments

Performance far below result with Raspberry Pi 3 B+
于老师您好！

我在用Raspberry Pi 3 B+做调试，看到项目README的结论，这款设备在320x240的图片下，单核性能是8.1fps，多核性能是23.74fps。但是我这边在320x240的实际测试结果，多核只有4.2fps，远远低于前面给出的结论。方便的话，请帮我找找原因，谢谢！

//下面是一些环境参数

os: 2020-02-13-raspbian-buster-lite

gcc version: 8.3.0 (Raspbian 8.3.0-6+rpi1) (本地编译)

cmake options: AVX512 = OFF AVX2 = OFF NEON = ON OpenMP = TRUE DEMO = ON add_compile_options(-mfpu=neon) (不加编译不通过)

opencv: 3.2.0

//cmake的输出 -- The C compiler identification is GNU 8.3.0 -- The CXX compiler identification is GNU 8.3.0 -- Check for working C compiler: /usr/bin/cc -- Check for working C compiler: /usr/bin/cc -- works -- Detecting C compiler ABI info -- Detecting C compiler ABI info - done -- Detecting C compile features -- Detecting C compile features - done -- Check for working CXX compiler: /usr/bin/c++ -- Check for working CXX compiler: /usr/bin/c++ -- works -- Detecting CXX compiler ABI info -- Detecting CXX compiler ABI info - done -- Detecting CXX compile features -- Detecting CXX compile features - done fatal: No names found, cannot describe anything. BUILD_VERSION:v0.0.1 Using ENON -- Found OpenMP_C: -fopenmp (found version "4.5") -- Found OpenMP_CXX: -fopenmp (found version "4.5") -- Found OpenMP: TRUE (found version "4.5")
-- Performing Test COMPILER_HAS_HIDDEN_VISIBILITY -- Performing Test COMPILER_HAS_HIDDEN_VISIBILITY - Success -- Performing Test COMPILER_HAS_HIDDEN_INLINE_VISIBILITY -- Performing Test COMPILER_HAS_HIDDEN_INLINE_VISIBILITY - Success -- Performing Test COMPILER_HAS_DEPRECATED_ATTR -- Performing Test COMPILER_HAS_DEPRECATED_ATTR - Success -- Found OpenCV: /usr (found version "3.2.0") AVX512 = OFF AVX2 = OFF NEON = ON OpenMP = TRUE DEMO = ON -- Configuring done -- Generating done -- Build files have been written to: /home/pi/dev/libfacedetection/build

//benchmark的运行结果 [email protected]:~/dev/libfacedetection/build $ ./benchmark 320_0.jpg There are 4 threads, 4 processors. cnn facedetection average time = 238.36ms | 4.20 FPS
opened by iamliuyin 19

运行速度比预期慢很多

于老师，您好！

我在windows系统运行图片人脸检测，912*524分辨率的png图片，可以正常完整运行，但是facedetect_cnn 这个函数耗时1903ms，我知道这是不合理的，但是我没有找到原因。电脑配置不差。使用默认的cmakeLists编译的facedetection_export.h，使用detect-image.cpp的示例代码做了一点修改，把图片地址写在代码里的。请问是会是什么原因导致？

opened by johniao 17

face detection with thermal images?

Do you have any activity for using your algorithm for detecting face on thermal images? It is very interesting now especially by Corona, where automatical facial temperature detection

opened by Thanh-Binh 11
读取位置 0xFFFFFFFFFFFFFFFF 时发生访问冲突

您好，我用VS2013在x86和x64环境下没有错误提示，但运行示例时libfacedetectcnn-example都出现以下错误提示，请大家帮忙指点：一直错误提示： 0x000000013F659557 处的第一机会异常(在 LibFaceDetection.exe 中): 0xC0000005: 读取位置 0xFFFFFFFFFFFFFFFF 时发生访问冲突。

代码停留在： _mm256_store_si256((__m256i*)sumarray, sum_int32x8);

没有更改过源文件。

opened by yuanYangSee 11
Just ported to android and NOT run as fast as it told,can anyone figure out what is wrong with it?

I ported libfacedetection with jni and build with ndk19, then ran on a Nokia 7 plus android phone which cpu is Qualcomm 660. The face rect is given correctly but took almost 700ms to get the result without couting the convert time of bitmap to rgb. Barely can it run 2fps... And last time I check ,NEON was enabled. Can anyone help me with it ?

opened by dpmaycry 11

Can not alloc memory for data.

I use your library in java with jni. And this error happens if i'm run "facedetect_frontal" function about 350-400 times. Can you tell me what mean this error ? How i can resolve it ? Thank you. ` #include "faceSwap_OurNativeLib.h" #include "facedetect-dll.h" #include <opencv2/opencv.hpp> #include #include #include using namespace std; using namespace cv; #define DETECT_BUFFER_SIZE 0x20000

JNIEXPORT jintArray JNICALL Java_faceswap_OurNativeLib_getPointsAndRec (JNIEnv * env, jobject obj, jlong matAddr, jfloat scale, jint min_neighbors,jint minObjectWidth,jint maxObjectWidth,jint dolandmark) { jintArray rezult = nullptr; Mat * inputMat = (Mat*)matAddr; Mat gray; cvtColor(*inputMat, gray, CV_BGR2GRAY); int * pResults = NULL; //pBuffer is used in the detection functions. //If you call functions in multiple threads, please create one buffer for each thread! unsigned char * pBuffer = (unsigned char *)malloc(DETECT_BUFFER_SIZE); if (!pBuffer) { fprintf(stderr, "Can not alloc buffer.\n"); rezult = env->NewIntArray(1); return rezult; }

///////////////////////////////////////////
// frontal face detection / 68 landmark detection
// it's fast, but cannot detect side view faces
//////////////////////////////////////////
//!!! The input image must be a gray one (single-channel)
//!!! DO NOT RELEASE pResults !!!
pResults = facedetect_frontal(pBuffer, (unsigned char*)(gray.ptr(0)), gray.cols, gray.rows, (int)gray.step,
	scale, min_neighbors, minObjectWidth, maxObjectWidth, dolandmark);


if ((*pResults) <= 0)
{
	fprintf(stderr, "Nofaces.\n");
	rezult = env->NewIntArray(1);
	return rezult;
}

rezult = env->NewIntArray((*pResults) * 140);
if (rezult == nullptr)
{
	fprintf(stderr, "Can not alloc rezult memory.\n");
	rezult = env->NewIntArray(1);
	return rezult;
}
//jint bodyOfrezult[140];
//for (int i = 0; i < (pResults ? *pResults : 0); i++)
//{
//	short * p = ((short*)(pResults + 1)) + 142 * i;
//	for (int k = 0; k < 4; k++)
//	{
//		bodyOfrezult[k] = p[k];
//	}
//	if (dolandmark)
//	{
//		for (int j = 4; j < 140; j++)
//		{
//			bodyOfrezult[j] = ((int)p[2 + j]);
//		}
//		env->SetIntArrayRegion(rezult, i * 140, 140 , bodyOfrezult);
//	}
//}
free(pBuffer);
gray.release();
return rezult;

}

opened by ussernamenikita 11

🔥🔥🔥 [Guide] How to use this lib in iOS

This lib is very useful! I try it in iOS and successful run.

download or clone this lib in your computer;
create New Xcode project;
add this lib's src to your project;
add the system lib libc++.tbd and other you need framework(eg. ACFoundation.framework etc.);
download opencv2.framework and add it in your project;
follow the example file to write the code.

!!!

modify facedetectcnn.h

//#define _ENABLE_AVX2 //Please enable it if X64 CPU
#define _ENABLE_NEON //Please enable it if ARM CPU

modify .m to .mm
import lib in your .mm

#import <opencv2/opencv.hpp>
#import <opencv2/imgcodecs/ios.h>
#import "ViewController.h"
#import "facedetectcnn.h"

you must import the opencv2/opencv.hpp first !!!

MyCode:

#import <opencv2/opencv.hpp>
#import <opencv2/imgcodecs/ios.h>
#import "ViewController.h"
#import "facedetectcnn.h"

//define the buffer size. Do not change the size!
#define DETECT_BUFFER_SIZE 0x20000
using namespace cv;

@implementation ViewController

- (UIImage *)loadImageAndDectect:(const char *)image_file{
    Mat img = imread(image_file);
    if (img.empty()) {
        fprintf(stderr, "Can not load the image file %s.\n", image_file);
        return nil;
    }
    
    int *pResults = NULL;
    unsigned char * pBuffer = (unsigned char *)malloc(DETECT_BUFFER_SIZE);
    if (!pBuffer) {
        fprintf(stderr, "Can not alloc buffer.\n");
        return nil;
    }
    pResults = facedetect_cnn(pBuffer, (unsigned char *)(img.ptr(0)), img.cols, img.rows, (int)img.step);
    printf("%d faces detected.\n", (pResults ? *pResults : 0));
    Mat result_cnn = img.clone();;
    //print the detection results
    for(int i = 0; i < (pResults ? *pResults : 0); i++)
    {
        short * p = ((short*)(pResults+1))+142*i;
        int x = p[0];
        int y = p[1];
        int w = p[2];
        int h = p[3];
        int neighbors = p[4];
        int angle = p[5];
        
        printf("face_rect=[%d, %d, %d, %d], neighbors=%d, angle=%d\n", x,y,w,h,neighbors, angle);
        rectangle(result_cnn, cv::Rect(x, y, w, h), Scalar(0, 255, 0), 2);
    }
    
    free(pBuffer);
    
    return MatToUIImage(result_cnn);
}


- (void)viewDidLoad {
    [super viewDidLoad];
    // Do any additional setup after loading the view, typically from a nib.
    
    UIImageView *imageView = [[UIImageView alloc] initWithFrame:[UIScreen mainScreen].bounds];
    imageView.contentMode = UIViewContentModeScaleAspectFit;
    [self.view addSubview:imageView];
    
    NSString *path = [[NSBundle mainBundle] pathForResource:@"test" ofType:@".jpg"];
    
    imageView.image = [self loadImageAndDectect:[path UTF8String]];
}


@end

opened by zycslog 10

CMakeLists.txt doesn't work

[email protected]:~/Projects/libfacedetection/build$ make -j4
[ 16%] Building CXX object CMakeFiles/demo.dir/src/facedetectcnn.cpp.o
/home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp: In function ‘float dotProductFloatChGeneral(float*, float*, int, int)’:
/home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:115:39: warning: AVX vector return without AVX enabled changes the ABI [-Wpsabi]
     __m256 sumvec = _mm256_setzero_ps();
                                       ^
In file included from /usr/lib/gcc/x86_64-linux-gnu/7/include/immintrin.h:41:0,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.h:54,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:39:
/usr/lib/gcc/x86_64-linux-gnu/7/include/avxintrin.h:1227:1: error: inlining failed in call to always_inline ‘__m256 _mm256_setzero_ps()’: target specific option mismatch
 _mm256_setzero_ps (void)
 ^~~~~~~~~~~~~~~~~
/home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:115:38: note: called from here
     __m256 sumvec = _mm256_setzero_ps();
                     ~~~~~~~~~~~~~~~~~^~
In file included from /usr/lib/gcc/x86_64-linux-gnu/7/include/immintrin.h:85:0,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.h:54,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:39:
/usr/lib/gcc/x86_64-linux-gnu/7/include/fmaintrin.h:63:1: error: inlining failed in call to always_inline ‘__m256 _mm256_fmadd_ps(__m256, __m256, __m256)’: target specific option mismatch
 _mm256_fmadd_ps (__m256 __A, __m256 __B, __m256 __C)
 ^~~~~~~~~~~~~~~
/home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:127:33: note: called from here
         sumvec = _mm256_fmadd_ps(avec, bvec, sumvec);
                  ~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~
In file included from /usr/lib/gcc/x86_64-linux-gnu/7/include/immintrin.h:41:0,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.h:54,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:39:
/usr/lib/gcc/x86_64-linux-gnu/7/include/avxintrin.h:871:1: error: inlining failed in call to always_inline ‘__m256 _mm256_load_ps(const float*)’: target specific option mismatch
 _mm256_load_ps (float const *__P)
 ^~~~~~~~~~~~~~
/home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:120:14: note: called from here
         bvec = _mm256_load_ps(p2 + i);
         ~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
In file included from /usr/lib/gcc/x86_64-linux-gnu/7/include/immintrin.h:41:0,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.h:54,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:39:
/usr/lib/gcc/x86_64-linux-gnu/7/include/avxintrin.h:871:1: error: inlining failed in call to always_inline ‘__m256 _mm256_load_ps(const float*)’: target specific option mismatch
 _mm256_load_ps (float const *__P)
 ^~~~~~~~~~~~~~
/home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:119:14: note: called from here
         avec = _mm256_load_ps(p1 + i);
         ~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
In file included from /usr/lib/gcc/x86_64-linux-gnu/7/include/immintrin.h:41:0,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.h:54,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:39:
/usr/lib/gcc/x86_64-linux-gnu/7/include/avxintrin.h:269:1: error: inlining failed in call to always_inline ‘__m256 _mm256_hadd_ps(__m256, __m256)’: target specific option mismatch
 _mm256_hadd_ps (__m256 __X, __m256 __Y)
 ^~~~~~~~~~~~~~
/home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:131:28: note: called from here
     sumvec = _mm256_hadd_ps(sumvec, sumvec);
              ~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~
In file included from /usr/lib/gcc/x86_64-linux-gnu/7/include/immintrin.h:41:0,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.h:54,
                 from /home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:39:
/usr/lib/gcc/x86_64-linux-gnu/7/include/avxintrin.h:269:1: error: inlining failed in call to always_inline ‘__m256 _mm256_hadd_ps(__m256, __m256)’: target specific option mismatch
 _mm256_hadd_ps (__m256 __X, __m256 __Y)
 ^~~~~~~~~~~~~~
/home/lvmc/Projects/libfacedetection/src/facedetectcnn.cpp:132:28: note: called from here
     sumvec = _mm256_hadd_ps(sumvec, sumvec);
              ~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~
CMakeFiles/demo.dir/build.make:134: recipe for target 'CMakeFiles/demo.dir/src/facedetectcnn.cpp.o' failed
make[2]: *** [CMakeFiles/demo.dir/src/facedetectcnn.cpp.o] Error 1
CMakeFiles/Makefile2:67: recipe for target 'CMakeFiles/demo.dir/all' failed
make[1]: *** [CMakeFiles/demo.dir/all] Error 2
Makefile:83: recipe for target 'all' failed
make: *** [all] Error 2

opened by Grabber 9

undefined reference to `facedetect_cnn(unsigned char*, unsigned char*, int, int, int)'

我在Ubantu上运行g++ libfacedetectcnn-example.cpp ./test.jpg -O3 pkg-config --libs --cflags opencv4 undefined reference to `facedetect_cnn(unsigned char*, unsigned char*, int, int, int)'，我已经加头文件了啊

opened by Adhders 9

How to Use Android Platform??

Why is it that after I migrated to Android, the recognition time of `keliamoniz1.jpg` in the sample image was 2800 milliseconds?

This is my CMakeList.txt:

cmake_minimum_required(VERSION 3.4.1)

SET(CMAKE_BUILD_TYPE Release)

set(CMAKE_SYSTEM_NAME Linux)
set(CMAKE_SYSTEM_VERSION 1)
set(CMAKE_SYSTEM_PROCESSOR "aarch64")

add_definitions(-O3)

add_definitions(-D_ENABLE_INT8)
#add_definitions(-D_ENABLE_AVX2)

SET(CMAKE_C_FLAGS "$ENV{CFLAGS} -O3 -Wall")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=gnu++11 -O3 -pthread")
SET(CMAKE_CXX_FLAGS_RELEASE "$ENV{CXXFLAGS} -O3 -Wall")

#set(CMAKE_C_COMPILER_ABI armeabi-v7a)

set(OPENCV_PATH /home/ww/Downloads/OpenCV-android-sdk)
set(CMAKE_VERBOSE_MAKEFILE on)

include_directories(${OPENCV_PATH}/sdk/native/jni/include)
add_library(lib_opencv STATIC IMPORTED)
set_target_properties(lib_opencv
        PROPERTIES
        IMPORTED_LOCATION ${PROJECT_SOURCE_DIR}/src/main/jniLibs/${ANDROID_ABI}/libopencv_java4.so
        )

add_library( 
        native-lib

        SHARED

        src/main/cpp/native-lib.cpp
        src/main/cpp/facedetectcnn.cpp
        src/main/cpp/facedetectcnn-floatdata.cpp
        src/main/cpp/facedetectcnn-int8data.cpp
        src/main/cpp/facedetectcnn-model.cpp
        )

find_library( 
        log-lib
        log)

target_link_libraries( 
        native-lib
        ${log-lib}
        lib_opencv
        )

Android device CPU is : mtk6737

opened by xiaomochen520 8

多线程调用

您好，11 月 10 日更新的这个版本您说的是可以在多线程调用，但是我用 VS2015（以及 opencv 3.1）测试发现，还是没法多线程同时调用啊，必须要加锁。另外，您提供的 dll 似乎是 VS2010 编译的，而新版的接口里面用到了 vector，编译器版本不匹配时在 vector 析构的时候就会出现问题，除非调用之前为 vector 预留足够的空间。

opened by sean028 8
关于facedetect_cnn函数定义

我在facedetectcnn-model.cpp中找到了facedetect_cnn函数的定义，注释写着“input image, it must be RGB (three-channel) image!”而在头文件中facedetectcnn.h写着"input image, it must be BGR (three channels) insteed of RGB image!"。我觉得虽然是个小问题，但不一致的代码注释可能引起阅读混乱？

opened by biepenghaomie 0
[MSVC] Libfacedetection build failed due to error C2039: 'to_string': is not a member of 'std'
Description: Recently, we updated the commit of bitcoin for MSVC RWC testing, it failed to build due to the error C2039 and error C3861 like below, could you please help take a look? Thanks.

F:\gitP\ShiqiYu\libfacedetection\src\facedetectcnn.cpp(638,17): error C2039: 'to_string': is not a member of 'std' [F:\gitP\ShiqiYu\libfacedetection\build_amd64\facedetection.vcxproj] F:\gitP\ShiqiYu\libfacedetection\src\facedetectcnn.cpp(638,26): error C3861: 'to_string': identifier not found [F:\gitP\ShiqiYu\libfacedetection\build_amd64\facedetection.vcxproj]

Expected behavior: Build successfully.

Actual behavior: Build failed with error C2039 and error C3861

To reprodece:

open VS2019 x64 Native Tools command.

git clone https://github.com/ShiqiYu/libfacedetection F:\gitP\ShiqiYu\libfacedetection

mkdir F:\gitP\ShiqiYu\libfacedetection\build_amd64 and cd F:\gitP\ShiqiYu\libfacedetection\build_amd64

cmake -G "Visual Studio 16 2019" -A x64 -DCMAKE_SYSTEM_VERSION=10.0.18362.0 -DCMAKE_INSTALL_PREFIX=install -DBUILD_SHARED_LIBS=ON -DCMAKE_BUILD_TYPE=Release -DDEMO=OFF ..

msbuild /m /p:Platform=x64 /p:Configuration=Release libfacedetection.sln /t:Rebuild

More info: The commit of Libfacedetection we use: 4092665 VS version: VS2019 (v16.11.20)

Detailed log: build.log
opened by Zhaojun-Liu 2
Is 150ms normal for AMD Ryzen 5 3550H in Single-thread？
I ran the detect-camera.cpp in example, and the average time consumption of face detection is 150ms. The environment is:

AMD Ryzen 5 3550H

16GB RAM

Microsoft Visual Studio 2022

By the way, I have enabled AVX2 and -O2 optimization when compling.
opened by YjFnajEz 3
Problem about FDDB score

教授您好，有鉴于您的测试性能与精度相当地高，我使用您的模型测试FDDB的结果如下(红框部分)：

我想请问教授这个测试结果是正常的吗，我有看到有些评测说您的模型可达到 0.994 平均精度(AP)，但我的测试都只达到0.968左右，

我目前是个程式菜鸟，有些地方不太了解其中的意义，如果有什么地方有做错或少做的话请您给与指教，谢谢！

opened by asiagodtonegg3beo 2
Upside-down images

Good morning. How can I implement processing of upside-down images? Looks like keypoints are the same, so I can't just check them and rotate image. Is there any tool or hint to work with that images?

Thank you.

opened by snolentinque 1

Releases(v3.0)

v3.0(Sep 24, 2021)

Depth-wise and point-wise convolutional layers for the backbone.

Performance on WIDER Face: AP_easy=0.856, AP_medium=0.842, AP_hard=0.727
Source code(tar.gz)
Source code(zip)
v2.0(Feb 9, 2021)

SSD structure network for face detection and 5 facial landmarks
Source code(tar.gz)
Source code(zip)

Owner

Shiqi Yu

Associate Professor, Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen, China.

GitHub Repository

SHRIMP: Sparser Random Feature Models via Iterative Magnitude Pruning

SHRIMP: Sparser Random Feature Models via Iterative Magnitude Pruning This repository is the official implementation of "SHRIMP: Sparser Random Featur

0 Dec 16, 2021

Training PSPNet in Tensorflow. Reproduce the performance from the paper.

Training Reproduce of PSPNet. (Updated 2021/04/09. Authors of PSPNet have provided a Pytorch implementation for PSPNet and their new work with support

126 Jul 13, 2022

A fast MoE impl for PyTorch

An easy-to-use and efficient system to support the Mixture of Experts (MoE) model for PyTorch.

873 Jan 09, 2023

YOLOv5 detection interface - PyQt5 implementation

所有代码已上传，直接clone后，运行yolo_win.py即可开启界面。 2021/9/29：加入置信度选择界面是在ultralytics的yolov5基础上建立的，界面使用pyqt5实现，内容较简单，娱乐而已。功能：模型选择本地文件选择(视频图片均可) 开关摄像头

487 Dec 27, 2022

Qimera: Data-free Quantization with Synthetic Boundary Supporting Samples

Qimera: Data-free Quantization with Synthetic Boundary Supporting Samples This repository is the official implementation of paper [Qimera: Data-free Q

21 Nov 03, 2022

Python scripts for performing 3D human pose estimation using the Mobile Human Pose model in ONNX.

99 Dec 31, 2022

Boosted CVaR Classification (NeurIPS 2021)

Boosted CVaR Classification Runtian Zhai, Chen Dan, Arun Sai Suggala, Zico Kolter, Pradeep Ravikumar NeurIPS 2021 Table of Contents Quick Start Train

4 Feb 15, 2022

Pytorch implementation of CoCon: A Self-Supervised Approach for Controlled Text Generation

COCON_ICLR2021 This is our Pytorch implementation of COCON. CoCon: A Self-Supervised Approach for Controlled Text Generation (ICLR 2021) Alvin Chan, Y

79 Dec 18, 2022

A curated list of awesome resources combining Transformers with Neural Architecture Search

173 Jan 03, 2023

MediaPipe is a an open-source framework from Google for building multimodal

MediaPipe is a an open-source framework from Google for building multimodal (eg. video, audio, any time series data), cross platform (i.e Android, iOS, web, edge devices) applied ML pipelines. It is

3 Sep 30, 2022

Using VapourSynth with super resolution models and speeding them up with TensorRT.

VSGAN-tensorrt-docker Using image super resolution models with vapoursynth and speeding them up with TensorRT. Using NVIDIA/Torch-TensorRT combined wi

111 Jan 05, 2023

[NeurIPS 2021] "G-PATE: Scalable Differentially Private Data Generator via Private Aggregation of Teacher Discriminators"

G-PATE This is the official code base for our NeurIPS 2021 paper: "G-PATE: Scalable Differentially Private Data Generator via Private Aggregation of T

14 Oct 12, 2022

Notepy is a full-featured Notepad Python app

Notepy A full featured python text-editor Notable features Autocompletion for parenthesis and quote Auto identation Syntax highlighting Compile and ru

11 Sep 28, 2022

Ludwig Benchmarking Toolkit

Ludwig Benchmarking Toolkit The Ludwig Benchmarking Toolkit is a personalized benchmarking toolkit for running end-to-end benchmark studies across an

17 Nov 18, 2022

Omniverse sample scripts - A guide for developing with Python scripts on NVIDIA Ominverse

Omniverse sample scripts ここでは、NVIDIA Omniverse ( https://www.nvidia.com/ja-jp/om

37 Nov 17, 2022

The code for Expectation-Maximization Attention Networks for Semantic Segmentation (ICCV'2019 Oral)

EMANet News The bug in loading the pretrained model is now fixed. I have updated the .pth. To use it, download it again. EMANet-101 gets 80.99 on the

663 Nov 30, 2022

QQ Browser 2021 AI Algorithm Competition Track 1 1st Place Program

249 Jan 03, 2023

S2-BNN: Bridging the Gap Between Self-Supervised Real and 1-bit Neural Networks via Guided Distribution Calibration (CVPR 2021)

S2-BNN (Self-supervised Binary Neural Networks Using Distillation Loss) This is the official pytorch implementation of our paper: "S2-BNN: Bridging th

52 Dec 24, 2022

Deep learning models for change detection of remote sensing images

Change Detection Models (Remote Sensing) Python library with Neural Networks for Change Detection based on PyTorch. ⚡ ⚡ ⚡ I am trying to build this pr

176 Dec 24, 2022

CAMoE + Dual SoftMax Loss (DSL): Improving Video-Text Retrieval by Multi-Stream Corpus Alignment and Dual Softmax Loss

CAMoE + Dual SoftMax Loss (DSL): Improving Video-Text Retrieval by Multi-Stream Corpus Alignment and Dual Softmax Loss This is official implement of "

87 Dec 24, 2022