from arcsoft import lib_func from arcsoft.struct_info import * MOK = 0 # 函数调用成功状态码 MERR_ASF_ALREADY_ACTIVATED = 90114 # 已激活状态返回码 ASF_NONE = 0x00000000 # 无属性 ASF_FACE_DETECT = 0x00000001 # 此处detect可以是tracking或者detection两个引擎之一，具体的选择由detect mode 确定 ASF_FACERECOGNITION = 0x00000004 # 人脸特征 ASF_AGE = 0x00000008 # 年龄 ASF_GENDER = 0x00000010 # 性别 ASF_FACE3DANGLE = 0x00000020 # 3D角度 ASF_LIVENESS = 0x00000080 # RGB活体 ASF_IR_LIVENESS = 0x00000400 # 红外活体 ASF_DETECT_MODE_VIDEO = 0x00000000 # Video模式，一般用于多帧连续检测 ASF_DETECT_MODE_IMAGE = 0xFFFFFFFF # Image模式，一般用于静态图的单次检测 ASVL_PAF_RGB24_B8G8R8 = 0x201 # RGB24位图，bgr格式(opencv 读取图像的默认格式) ASVL_PAF_GRAY = 0x701 # 灰度图 # 人脸检测方向 ASF_OP_0_ONLY = 0x1 # 仅检测 0 度 ASF_OP_90_ONLY = 0x2 # 仅检测 90 度 ASF_OP_270_ONLY = 0x3 # 仅检测 270 度 ASF_OP_180_ONLY = 0x4 # 仅检测 180 度 ASF_OP_0_HIGHER_EXT = 0x5 # 全角度检测 # 检测到的人脸角度 ASF_OC_0 = 0x1 # 0 度 ASF_OC_90 = 0x2 # 90 度 ASF_OC_270 = 0x3 # 270 度 ASF_OC_180 = 0x4 # 180 度 ASF_OC_30 = 0x5 # 30 度 ASF_OC_60 = 0x6 # 60 度 ASF_OC_120 = 0x7 # 120 度 ASF_OC_150 = 0x8 # 150 度 ASF_OC_210 = 0x9 # 210 度 ASF_OC_240 = 0xa # 240 度 ASF_OC_300 = 0xb # 300 度 ASF_OC_330 = 0xc # 330 度 def ASFOnlineActivation(Appkey, SDKey): """ 在线激活 SDK :param Appkey: 官网获取的 APPID :param SDKey: 官网获取的 SDKKEY :return: 状态码 """ res = lib_func.ASFOnlineActivation(Appkey, SDKey) if res != 90114 and res != 0: print("激活失败!错误码:{}".format(res)) return res def ASFGetActiveFileInfo(): """ 获取激活文件信息 :return: 状态码， 激活文件信息 """ activeFileInfo = ASF_ActiveFileInfo() return lib_func.ASFGetActiveFileInfo(byref(activeFileInfo)), activeFileInfo class ArcFace(): def __init__(self): self.Handle = c_void_p() # 引擎指针对象 def ASFInitEngine(self, detectMode: int, detectFaceOrientPriority: int, detectFaceScaleVal: int, detectFaceMaxNum: int, combinedMask: int): """ 初始化引擎 :param detectMode: VIDEO 模式/IMAGE 模式 ,VIDEO 模式:处理连续帧的图像数据 IMAGE 模式:处理单张的图像数据 :param detectFaceOrientPriority: 人脸检测角度，推荐单一角度检测；IMAGE 模式下不支持全角度（ASF_OP_0_HIGHER_EXT）检测 :param detectFaceScaleVal: 识别的最小人脸比例（图片长边与人脸框长边的比值）,VIDEO 模式取值范围[2,32]，推荐值为 16 ,IMAGE 模式取值范围[2,32]，推荐值为 30 :param detectFaceMaxNum: 最大需要检测的人脸个数，取值范围[1,50] :param combinedMask: 需要启用的功能组合，可多选 :return: 状态码 """ return lib_func.ASFInitEngine(detectMode, detectFaceOrientPriority, detectFaceScaleVal, detectFaceMaxNum, combinedMask, byref(self.Handle)) def ASFDetectFaces(self, frame): """ 人脸检测 :param frame: 原始图像：注意：图片宽度必须 为 4 的倍数 :return: 状态码,人脸检测信息 """ height, width = frame.shape[:2] detectedFaces = ASF_MultiFaceInfo() res = lib_func.ASFDetectFaces(self.Handle, int(width), int(height), ASVL_PAF_RGB24_B8G8R8, frame.ctypes.data_as(POINTER(c_ubyte)), byref(detectedFaces)) return res, detectedFaces def ASFProcess_IR(self, frame, detectedFaces, processIRMask=ASF_IR_LIVENESS): """ :param frame: 原始图像：注意：图片宽度必须 为 4 的倍数 :param detectedFaces: 多人脸检测信息对象 :param processIRMask: 当前只有ASF_IR_LIVENESS 一种选择 注：检测的属性须在引擎初始化接口的 combinedMask 参 数中启用 :return: """ height, width = frame.shape[:2] res = lib_func.ASFProcess_IR(self.Handle, int(width), int(height), ASVL_PAF_GRAY, frame.ctypes.data_as(POINTER(c_ubyte)), byref(detectedFaces), processIRMask) return res def ASFFaceFeatureExtract(self, frame, singleFaceInfo: ASF_SingleFaceInfo): """ 人脸特征提取 :param frame: 原始图像：注意：图片宽度必须 为 4 的倍数 :param singleFaceInfo: 单个人脸检测框信息 :return: 状态码,人脸检测信息 """ height, width = frame.shape[:2] face_feature = ASF_FaceFeature() res = lib_func.ASFFaceFeatureExtract(self.Handle, int(width), int(height), ASVL_PAF_RGB24_B8G8R8, frame.ctypes.data_as(POINTER(c_ubyte)), singleFaceInfo, byref(face_feature)) copy_face_feature = ASF_FaceFeature() copy_face_feature.featureSize = face_feature.featureSize copy_face_feature.feature = lib_func.malloc(face_feature.featureSize) lib_func.memcpy(copy_face_feature.feature, face_feature.feature, face_feature.featureSize) return res, copy_face_feature def ASFFaceFeatureCompare(self, face_feature1: ASF_FaceFeature, face_feature2: ASF_FaceFeature): """ 人脸特征比较 :param face_feature1: 特征对象1 :param face_feature2: 特征对象2 :return: 状态码，人脸得分 """ compare_score = c_float() ret = lib_func.ASFFaceFeatureCompare(self.Handle, face_feature1, face_feature2, byref(compare_score)) return ret, compare_score.value def ASFProcess(self, frame, detectedFaces: ASF_MultiFaceInfo, combinedMask: int): """ 人脸信息检测（年龄/性别/人脸 3D 角度/rgb活体），最多支持 4 张人脸信息检测，超过部分返回未知（活体仅支持单张人脸检测，超出返回未知） :param frame: 原始图像：注意：图片宽度必须 为 4 的倍数 :param detectedFaces: 多人脸检测信息对象 :param combinedMask: 检测的属性（ASF_AGE、ASF_GENDER、ASF_FACE3DANGLE、ASF_LIVENESS），支持多选 注：检测的属性须在引擎初始化接口的 combinedMask 参 数中启用 :return: """ height, width = frame.shape[:2] res = lib_func.ASFProcess(self.Handle, int(width), int(height), ASVL_PAF_RGB24_B8G8R8, frame.ctypes.data_as(POINTER(c_ubyte)), byref(detectedFaces), combinedMask) return res def ASFGetFace3DAngle(self): """ 获取 3D 角度信息 :return: 状态码，人脸3d角度信息 """ angleInfo = ASF_Face3DAngle() res = lib_func.ASFGetFace3DAngle(self.Handle, angleInfo) return res, angleInfo def ASFGetLivenessScore(self): """ 获取 RGB 活体信息 :return:状态码，RGB 活体信息 """ rgbLivenessInfo = ASF_LivenessInfo() res = lib_func.ASFGetLivenessScore(self.Handle, byref(rgbLivenessInfo)) return res, rgbLivenessInfo def ASFGetAge(self): """ 获取年龄信息 :return: :状态码，年龄信息 """ ageInfo = ASF_AgeInfo() res = lib_func.ASFGetAge(self.Handle, byref(ageInfo)) return res, ageInfo def ASFGetGender(self): """ 获取性别信息 :return:状态码，性别信息 """ genderInfo = ASF_GenderInfo() res = lib_func.ASFGetGender(self.Handle, byref(genderInfo)) return res, genderInfo def ASFGetLivenessScore_IR(self): """ 获取IR识别结果 :return:状态码，IR识别结果 """ irLivenessInfo = ASF_LivenessInfo() res = lib_func.ASFGetLivenessScore_IR(self.Handle, byref(irLivenessInfo)) return res, irLivenessInfo def ASFSetLivenessParam(self, threshold: ASF_LivenessThreshold): """ 设置活体检测置信度 :return: """ return lib_func.ASFSetLivenessParam(self.Handle, threshold) def ASFUninitEngine(self): """ 销毁引擎 :return: 状态码 """ return lib_func.ASFUninitEngine(self.Handle) # 从二进制数据获取特征值 def ASFGetFeatureFromBinary(self, binaryValue): face_feature = ASF_FaceFeature() b = binaryValue face_feature.featureSize = b.__len__() face_feature.feature = lib_func.malloc(face_feature.featureSize) lib_func.memcpy(face_feature.feature, b, face_feature.featureSize) return face_feature