import sys import os import math import cv2 import dlib import numpy as np import Facial_Recognition_Render as fr import _pickle as cPickle import glob faceWidth = 320 faceHeight = 320 SKIP_FRAMES = 1 def alignFace(imFace, landmarks): l_x = landmarks[39][0] l_y = landmarks[39][1] r_x = landmarks[42][0] r_y = landmarks[42][1] dy = r_y - l_y dx = r_x - l_x # Convert from radians to degrees angle = math.atan2(dy, dx) * 180.0 / math.pi eyesCenter = ((l_x + r_x)*0.5, (l_y + r_y)*0.5) rotMatrix = cv2.getRotationMatrix2D(eyesCenter, angle, 1) alignedImFace = np.zeros(imFace.shape, dtype=np.uint8) alignedImFace = cv2.warpAffine(imFace, rotMatrix, (imFace.shape[1],imFace.shape[0])) return alignedImFace def face_detector_haarcascade(image): grey = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) resize_fx = 1 resize_fy = 1 grey = cv2.resize(grey, dsize=None, fx=resize_fx, fy=resize_fy, interpolation = cv2.INTER_AREA) pwd = sys.path[0] classfier = cv2.CascadeClassifier(pwd + "/Facial_models/haarcascade_frontalface_alt2.xml") faceRects = classfier.detectMultiScale(grey, scaleFactor=1.2, minNeighbors=1, minSize=(16, 16)) if len(faceRects) > 0: for faceRect in faceRects: x, y, w, h = faceRect x = int(x/resize_fx) y = int(y/resize_fy) w = int(w/resize_fx) h = int(h/resize_fy) cv2.rectangle(image, (x - 10, y - 10), (x + w + 10, y + h + 10), (0, 255, 0), 5) return image def face_detector_ssd(image): pwd = sys.path[0] net = cv2.dnn.readNetFromCaffe(pwd+"/Facial_models/deploy.prototxt", pwd+"/Facial_models/res10_300x300_ssd_iter_140000_fp16.caffemodel") resize = (300, 300) confidence_thres = 0.65 blob = cv2.dnn.blobFromImage(cv2.resize(image, dsize=resize), 1.0, resize, (104.0, 177.0, 123.0)) # blob = cv2.dnn.blobFromImage(cv2.resize(image, (300, 300)), 1.0, (300, 300), (104.0, 177.0, 123.0)) net.setInput(blob) detections = net.forward() h,w,c=image.shape for i in range(0, detections.shape[2]): confidence = detections[0, 0, i, 2] if confidence > confidence_thres: box = detections[0, 0, i, 3:7] * np.array([w, h, w, h]) (startX, startY, endX, endY) = box.astype("int") text = "{:.2f}%".format(confidence * 100) y = startY - 10 if startY - 10 > 10 else startY + 10 cv2.rectangle(image, (startX, startY), (endX, endY),(0, 255,0), 5) cv2.putText(image, text, (startX, y), cv2.FONT_HERSHEY_SIMPLEX, 1.00, (0, 255, 0), 3) return image def training_data_loader(): imagesFolder = sys.path[0]+"/Facial_images/face_rec/train/" subfolders = [] for x in os.listdir(imagesFolder): xpath = os.path.join(imagesFolder, x) if os.path.isdir(xpath): subfolders.append(xpath) imagePaths = [] labels = [] labelsMap = {} labelsMap[-1] = "unknown" for i, subfolder in enumerate(subfolders): labelsMap[i] = os.path.basename(subfolder) for x in os.listdir(subfolder): xpath = os.path.join(subfolder, x) if x.endswith('jpg') or x.endswith('pgm'): imagePaths.append(xpath) labels.append(i) imagesFaceTrain = [] labelsFaceTrain = [] faceDetector = dlib.get_frontal_face_detector() landmarkDetector = dlib.shape_predictor(sys.path[0]+"/Facial_models/shape_predictor_68_face_landmarks.dat") for j, imagePath in enumerate(imagePaths): im = cv2.imread(imagePath, 0) imHeight, imWidth = im.shape[:2] landmarks = fr.getLandmarks(faceDetector, landmarkDetector, im) landmarks = np.array(landmarks) if len(landmarks) == 68: x1Limit = landmarks[0][0] - (landmarks[36][0] - landmarks[0][0]) x2Limit = landmarks[16][0] + (landmarks[16][0] - landmarks[45][0]) y1Limit = landmarks[27][1] - 3*(landmarks[30][1] - landmarks[27][1]) y2Limit = landmarks[8][1] + (landmarks[30][1] - landmarks[29][1]) x1 = max(x1Limit,0) x2 = min(x2Limit, imWidth) y1 = max(y1Limit, 0) y2 = min(y2Limit, imHeight) imFace = im[y1:y2, x1:x2] alignedFace = alignFace(imFace, landmarks) alignedFace = cv2.resize(alignedFace, (faceHeight, faceWidth)) imagesFaceTrain.append(np.float32(alignedFace)/255.0) labelsFaceTrain.append(labels[j]) return imagesFaceTrain, labelsFaceTrain, labelsMap def training_recognizer(rec_type): imagesFaceTrain, labelsFaceTrain, labelsMap = training_data_loader() if (rec_type=='LBPH'): faceRecognizer = cv2.face.LBPHFaceRecognizer_create() print("Training using LBPH Faces") elif (rec_type=='Eigen'): faceRecognizer = cv2.face.EigenFaceRecognizer_create() print("Training using Eigen Faces") elif (rec_type=='Fisher'): faceRecognizer = cv2.face.FisherFaceRecognizer_create() print("Training using Fisher Faces") faceRecognizer.train(imagesFaceTrain, np.array(labelsFaceTrain)) faceRecognizer.write(sys.path[0]+'/Facial_models/face_rec_model.yml') with open(sys.path[0]+'/Facial_models/labels_map.pkl', 'wb') as f: cPickle.dump(labelsMap, f) def face_recognition_inference(rec_type): #testFiles = glob.glob(sys.path[0]+'/Facial_test_images/face_rec/test/*.jpg') #testFiles.sort() i = 0 correct = 0 error = 0 faceDetector = dlib.get_frontal_face_detector() landmarkDetector = dlib.shape_predictor(sys.path[0]+'/Facial_models/shape_predictor_68_face_landmarks.dat') if (rec_type=='LBPH'): faceRecognizer = cv2.face.LBPHFaceRecognizer_create() print("Test using LBPH Faces") elif (rec_type=='Eigen'): faceRecognizer = cv2.face.EigenFaceRecognizer_create() print("Test using Eigen Faces") elif (rec_type=='Fisher'): faceRecognizer = cv2.face.FisherFaceRecognizer_create() print("Test using Fisher Faces") faceRecognizer.read(sys.path[0]+'/Facial_models/face_rec_model.yml') labelsMap = np.load(sys.path[0]+'/Facial_models/labels_map.pkl', allow_pickle=True) cam = cv2.VideoCapture(1) while(True): #imagePath = testFiles[i] success, original = cam.read() im = cv2.resize(original, (640, 480)) i += 1 im = cv2.cvtColor(im,cv2.COLOR_BGR2GRAY) imHeight, imWidth = im.shape[:2] landmarks = fr.getLandmarks(faceDetector, landmarkDetector, im) landmarks = np.array(landmarks) if len(landmarks) == 68: x1Limit = landmarks[0][0] - (landmarks[36][0] - landmarks[0][0]) x2Limit = landmarks[16][0] + (landmarks[16][0] - landmarks[45][0]) y1Limit = landmarks[27][1] - 3*(landmarks[30][1] - landmarks[27][1]) y2Limit = landmarks[8][1] + (landmarks[30][1] - landmarks[29][1]) x1 = max(x1Limit,0) x2 = min(x2Limit, imWidth) y1 = max(y1Limit, 0) y2 = min(y2Limit, imHeight) imFace = im[y1:y2, x1:x2] alignedFace = alignFace(imFace, landmarks) alignedFace = cv2.resize(alignedFace, (faceHeight, faceWidth)) imFaceFloat = np.float32(alignedFace)/255.0 predictedLabel = -1 predictedLabel, score = faceRecognizer.predict(imFaceFloat) center = ( int((x1 + x2) /2), int((y1 + y2)/2) ) radius = int((y2-y1)/2.0) text = '{} {}%'.format(labelsMap[predictedLabel],round(score, 5)) cv2.rectangle(original, (x1, y1), (x2, y2), (0, 255, 0), 5) cv2.putText(original, text, (x1, y1-10), cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 0), 3) cv2.imshow('Face Recognition Demo', original) k = cv2.waitKey(10) cam.release() cv2.destroyAllWindows() if __name__=="__main__": mode = 'test' rec_type = 'Fisher' # 'LBPH' 'Fisher' 'Eigen' if mode == 'train': training_recognizer(rec_type) elif (mode == 'test'): face_recognition_inference(rec_type) # video process (keep it in case if needed) ''' cameraCapture = cv2.VideoCapture(1) success, frame = cameraCapture.read() while success and cv2.waitKey(1) == -1: success, frame = cameraCapture.read() face_detector_ssd(frame) cv2.imshow("video", frame) cameraCapture.release() cv2.destroyAllWindows() ''' # image process (keep it in case if needed) ''' image_name = "8.jpg" split_name = image_name.split(".") image_read_path = sys.path[0]+"/Facial_test_images/"+image_name image_save_path = sys.path[0]+"/Facial_test_images/output/"+split_name[0]+"_result."+split_name[1] image = cv2.imread(image_read_path) image = face_detector_ssd(image) image = face_detector_haarcascade(image) print(image_save_path) cv2.imwrite(image_save_path, image) cv2.imshow("result", image) cv2.waitKey() cv2.destroyAllWindows() '''