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

[Yak0xff]

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

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

!!!

1. modify facedetectcnn.h

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

2. modify .m to .mm
3. 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

[ShiqiYu]

If possible, could you submit a pull request with the iOS example and HowTo? I will merge your files submitted to the repository.

…

On Fri, Mar 15, 2019 at 11:08 AM Robin ***@***.***> wrote: This lib is very useful! I try it in iOS and successful run. 1. download or clone this lib in your computer; 2. create New Xcode project; 3. add this lib's src to your project; 4. add the system lib libc++.tbd and other you need framework(eg. ACFoundation.framework etc.); 5. download opencv2.framework and add it in your project; 6. follow the example file to write the code. *!!!* 1. modify facedetectcnn.h //#define _ENABLE_AVX2 //Please enable it if X64 CPU #define _ENABLE_NEON //Please enable it if ARM CPU 1. modify .m to .mm 2. 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 — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#90>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AG0eHZARHGfnU58IRciLvhUYPu87O-ICks5vWw6hgaJpZM4b1ndz> .

[ewen1024]

@RobinChao Do you have the running time on IOS. I tested it on iphone 6s and iphone x, much slower than the raspberry platform

[Yak0xff]

@ewen1024 Sorry, i didn't have the running time. Did you input huge image to NN, if you resize the image or use small size image, maybe running faster a little. Thanks.

[ewen1024]

Thx for your reply.
By enabling neon int8 and optimization -o3, I already got the running time in 100ms.
But I failed to enable openmp on ios which said the library only supports x86. Do you have any idea on this?\

[ShiqiYu]

Currently no OpenMP support in the source code. I will add it soon.

…

On Thu, Apr 25, 2019 at 11:37 AM ewen1024 ***@***.***> wrote: Thx for your reply. By enabling neon int8 and optimization -o3, I already got the running time in 100ms. But I failed to enable openmp on ios which said the library only supports x86. Do you have any idea on this?\ — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#90 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABWR4HKJONSUK2MG6RT4M33PSERPVANCNFSM4G6WO5ZQ> .

[ShiqiYu]

Sorry for the wrong reply. You can compile the source code with -fopenmp (GCC) to enable OpenMP.

…

On Thu, Apr 25, 2019 at 12:00 PM Shiqi Yu ***@***.***> wrote: Currently no OpenMP support in the source code. I will add it soon. On Thu, Apr 25, 2019 at 11:37 AM ewen1024 ***@***.***> wrote: > Thx for your reply. > By enabling neon int8 and optimization -o3, I already got the running > time in 100ms. > But I failed to enable openmp on ios which said the library only supports > x86. Do you have any idea on this?\ > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#90 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/ABWR4HKJONSUK2MG6RT4M33PSERPVANCNFSM4G6WO5ZQ> > . >

[ewen1024]

@ShiqiYu On ios platform, it seems the compiler clang doesn't support openmp.
So I can't compile the code with -fopenmp. Maybe I need to find a way to add support of openmp for the clang

[ShiqiYu]

You can change the macro _OPENMP in the source code to the clang supported code. According to my experiments, OpenMP cannot improve the speed obviously on Intel platform.

…

On Thu, Apr 25, 2019 at 12:09 PM ewen1024 ***@***.***> wrote: @ShiqiYu <https://github.com/ShiqiYu> On ios platform, it seems the compiler clang doesn't support openmp. So I can't compile the code with -fopenmp. Maybe I need to find a way to add support of openmp for the clang — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#90 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABWR4HMKS5K5HVOG2MLGEM3PSEVGBANCNFSM4G6WO5ZQ> .

[ewen1024]

@ShiqiYu appreciate for your help！

[zpfate]

请问一下为什么在ios上图片变色了, 或者使用opencv的ios的函数UIImageToMat, 无法识别出人脸