基於python、虹軟實現人臉檢測，人臉識別

時間 2019-11-19

原文原文鏈接

虹軟的人臉識別技術也是很強的，重要的是他免費提供了離線的sdk，還提供了實例，這個是目前幾家研究人臉識別的大公司裏面少有的。識別能力正經常使用仍是能夠的。我這個代碼是調用的離線sdk實現的app

from arcsoft import CLibrary, ASVL_COLOR_FORMAT, ASVLOFFSCREEN,c_ubyte_p,FaceInfo
from arcsoft.utils import BufferInfo, ImageLoader
from arcsoft.AFD_FSDKLibrary import *
from ctypes import *
import traceback
import cv2
import time
 
 
APPID = c_char_p(b'your id')
FD_SDKKEY = c_char_p(b'your key')
FD_WORKBUF_SIZE = 20 * 1024 * 1024
MAX_FACE_NUM = 50
bUseYUVFile = False
bUseBGRToEngine = True
 
def doFaceDetection(hFDEngine, inputImg):   #對圖像中的人臉進行定位
    faceInfo = []
 
    pFaceRes = POINTER(AFD_FSDK_FACERES)()
    ret = AFD_FSDK_StillImageFaceDetection(hFDEngine, byref(inputImg), byref(pFaceRes))
#ret 爲0
 
    if ret != 0:
        print(u'AFD_FSDK_StillImageFaceDetection 0x{0:x}'.format(ret))
        return faceInfo
    faceRes = pFaceRes.contents
    print('******')
 
    facecont=faceRes.nFace  #faceRes 是一個對象因此 輸出會是一個地址值 而他的一個屬性nface是表示的是人臉的個數
    print('%d 我的臉' %facecont)
 
 
    if faceRes.nFace > 0:
        for i in range(0, faceRes.nFace):
            rect = faceRes.rcFace[i]
            orient = faceRes.lfaceOrient[i]
            faceInfo.append(FaceInfo(rect.left,rect.top,rect.right,rect.bottom,orient))
 
 
    return faceInfo
 
def loadImage(filePath):
 
    inputImg = ASVLOFFSCREEN()
 
    if bUseBGRToEngine:  #true
        bufferInfo = ImageLoader.getBGRFromFile(filePath)
        inputImg.u32PixelArrayFormat = ASVL_COLOR_FORMAT.ASVL_PAF_RGB24_B8G8R8
        inputImg.i32Width = bufferInfo.width
        inputImg.i32Height = bufferInfo.height
        inputImg.pi32Pitch[0] = bufferInfo.width*3
        inputImg.ppu8Plane[0] = cast(bufferInfo.buffer, c_ubyte_p)
        inputImg.ppu8Plane[1] = cast(0, c_ubyte_p)
        inputImg.ppu8Plane[2] = cast(0, c_ubyte_p)
        inputImg.ppu8Plane[3] = cast(0, c_ubyte_p)
    else:
        bufferInfo = ImageLoader.getI420FromFile(filePath)
        inputImg.u32PixelArrayFormat = ASVL_COLOR_FORMAT.ASVL_PAF_I420
        inputImg.i32Width = bufferInfo.width
        inputImg.i32Height = bufferInfo.height
        inputImg.pi32Pitch[0] = inputImg.i32Width
        inputImg.pi32Pitch[1] = inputImg.i32Width // 2
        inputImg.pi32Pitch[2] = inputImg.i32Width // 2
        inputImg.ppu8Plane[0] = cast(bufferInfo.buffer, c_ubyte_p)
        inputImg.ppu8Plane[1] = cast(addressof(inputImg.ppu8Plane[0].contents) + (inputImg.pi32Pitch[0] * inputImg.i32Height), c_ubyte_p)
        inputImg.ppu8Plane[2] = cast(addressof(inputImg.ppu8Plane[1].contents) + (inputImg.pi32Pitch[1] * inputImg.i32Height // 2), c_ubyte_p)
        inputImg.ppu8Plane[3] = cast(0, c_ubyte_p)
    inputImg.gc_ppu8Plane0 = bufferInfo.buffer
 
    return inputImg
 
 
if __name__ == u'__main__':
    t=time.time()
    print(u'#####################################################')
 
    # init Engine
    pFDWorkMem = CLibrary.malloc(c_size_t(FD_WORKBUF_SIZE))
    hFDEngine = c_void_p()
    ret = AFD_FSDK_InitialFaceEngine(APPID, FD_SDKKEY, pFDWorkMem, c_int32(FD_WORKBUF_SIZE), byref(hFDEngine), AFD_FSDK_OPF_0_HIGHER_EXT, 32, MAX_FACE_NUM)
    #ret 爲0
    if ret != 0:
        CLibrary.free(pFDWorkMem)
        print(u'AFD_FSDK_InitialFaceEngine ret 0x{:x}'.format(ret))
        exit(0)
#--------------------------------以上部分兩個函數以及主函數的幾條語句不變-----------------------------------------------------------
 
    filePath = '001.jpg'
    inputImg = loadImage(filePath)  #調用loadImage函數  返回一種格式（目前還不知道這種格式是什麼）
 
    frame=cv2.imread(filePath)
    # do Face Detect
 
    faceInfos = doFaceDetection(hFDEngine, inputImg)  #調用dofaceDetection函數 進行圖像處理檢測人臉
    #print('faceInfos %s'% faceInfos[0])
 
    for i in range(0, len(faceInfos)):
        rect = faceInfos[i]
        print(u'{} ({} {} {} {}) orient {}'.format(i, rect.left, rect.top, rect.right, rect.bottom, rect.orient))
        cv2.rectangle(frame, (rect.left, rect.top), (rect.right, rect.bottom), (0, 0, 255), 2)
        cropimg=frame[rect.top:rect.bottom,rect.left:rect.right]# 使用opencv裁剪照片  把人臉的照片裁剪下來
        cv2.imwrite('crop-photo/'+str(i)+'.jpg',cropimg)  # 把人臉照片保存下來
 
 
 
 
 
    AFD_FSDK_UninitialFaceEngine(hFDEngine)     # release Engine
    cv2.imshow('tuxiang',frame)
    cv2.waitKey(1)
    print('所用時間爲{} '.format(time.time()-t)) #不進行保存圖片 0.12s  保存圖片0.16s
    time.sleep(1)
 
    CLibrary.free(pFDWorkMem)
    print(u'#####################################################')

運行結果函數