face recognition without api

api/face_rec.py

@@ -0,0 +1,69 @@
+# Face recognition function for api
+
+# Reference: Muhammad Ardi
+# from https://medium.com/@muhammad_ardi/a-simple-python-code-for-face-recognition-3266e9fad3be
+import os
+import cv2
+import face_recognition
+import numpy as np
+from tqdm import tqdm
+
+class Face:
+    def __init__(self, bounding_box, cropped_face, name, feature_vector):
+        self.bounding_box = bounding_box
+        self.cropped_face = cropped_face
+        self.name = name
+        self.feature_vector = feature_vector
+
+def load_image(path):
+    image = cv2.imread(path)
+    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+    return image
+
+def create_database(filenames, images):
+    faces = []
+    for filename, image in tqdm(zip(filenames, images), total=len(filenames)):
+        loc = face_recognition.face_locations(image, model='hog')[0]
+        vec = face_recognition.face_encodings(image, [loc], num_jitters=20)[0]
+
+        top, right, bottom, left = loc
+
+        cropped_face = image[top:bottom, left:right]
+
+        face = Face(bounding_box=loc, cropped_face=cropped_face, name=filename.split('.')[0], feature_vector=vec)
+        faces.append(face)
+
+    return faces
+
+def detect_faces(image_test, faces, threshold=0.5):
+    locs_test = face_recognition.face_locations(image_test, model='hog')
+    vecs_test = face_recognition.face_encodings(image_test, locs_test, num_jitters=1)
+
+    match = "false"
+
+    for loc_test, vec_test in zip(locs_test, vecs_test):
+        distances = []
+        for face in faces:
+            distance = face_recognition.face_distance([vec_test], face.feature_vector)
+            distances.append(distance)
+
+        if np.min(distances) < threshold:
+            match = "true"
+            break
+
+    return match
+
+filenames = os.listdir('templates')
+images = [load_image(f'templates/{filename}') for filename in filenames]
+faces = create_database(filenames, images)
+
+# My modification
+tests = os.listdir('tests')
+for test in tests:
+    if(".jpg" not in test and ".png" not in test):
+        continue
+    print(f'test: {test}')
+    image_test = load_image(f'tests/{test}')
+
+    name = detect_faces(image_test, faces, threshold=0.5)
+    print(f'Pridict name: {name}\n')

api/run.py

@@ -0,0 +1,39 @@
+# API for face recognition
+
+# Reference: 10程式中
+# from https://ithelp.ithome.com.tw/articles/10280422
+
+import numpy as np
+from face_rec import create_database, detect_faces
+import base64
+
+from flask import Flask, request, jsonify
+# from flask_cors import CORS
+
+app = Flask(__name__)
+# CORS(app)
+
+@app.route('/')
+def index():
+    return 'This is the API for face recognition'
+
+@app.route('/predict', methods=['POST'])
+def postInput():
+    # 取得前端傳過來的數值
+    insertValues = request.get_json()
+    image1 = base64.b64decode(insertValues['picture1'])
+    test_image = base64.b64decode(insertValues['picture2'])
+
+    filenames = []
+    filenames.append("image1")
+    standard_image = []
+    standard_image.append(image1)
+
+    faces = create_database(filenames, standard_image)
+
+    result = detect_faces(test_image, faces, threshold=0.4)
+
+    return jsonify({'result': str(result)})
+
+if __name__ == '__main__':
+    app.run(port=5000, debug=True)

face_reconition/Readme.md

@@ -0,0 +1,5 @@
+# This folder is for face recognition.
+
+## Usage
+* put the authorized image in /templates
+* put the images that need examination in /tests

face_reconition/face_rec.py

@@ -0,0 +1,132 @@
+# Testcode for face recognition
+
+# Reference: Muhammad Ardi
+# from https://medium.com/@muhammad_ardi/a-simple-python-code-for-face-recognition-3266e9fad3be
+import os
+import cv2
+import face_recognition
+import numpy as np
+from tqdm import tqdm
+# for showing image
+import matplotlib.pyplot as plt
+
+class Face:
+    def __init__(self, bounding_box, cropped_face, name, feature_vector):
+        self.bounding_box = bounding_box
+        self.cropped_face = cropped_face
+        self.name = name
+        self.feature_vector = feature_vector
+
+def load_image(path):
+    image = cv2.imread(path)
+    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+    return image
+
+def show_image(image):
+    plt.imshow(image)
+    plt.xticks([])
+    plt.yticks([])
+    plt.show()
+    return
+
+def create_database(filenames, images):
+    faces = []
+    for filename, image in tqdm(zip(filenames, images), total=len(filenames)):
+        loc = face_recognition.face_locations(image, model='hog')[0]
+        vec = face_recognition.face_encodings(image, [loc], num_jitters=20)[0]
+
+        top, right, bottom, left = loc
+
+        cropped_face = image[top:bottom, left:right]
+
+        face = Face(bounding_box=loc, cropped_face=cropped_face, name=filename.split('.')[0], feature_vector=vec)
+        faces.append(face)
+
+    return faces
+
+def detect_faces(image_test, faces, threshold=0.6):
+    locs_test = face_recognition.face_locations(image_test, model='hog')
+    vecs_test = face_recognition.face_encodings(image_test, locs_test, num_jitters=1)
+
+    for loc_test, vec_test in zip(locs_test, vecs_test):
+        distances = []
+        for face in faces:
+            distance = face_recognition.face_distance([vec_test], face.feature_vector)
+            distances.append(distance)
+
+        if np.min(distances) > threshold:
+            pred_name = 'unknown'
+            print('Not match')
+            print('min distance:', np.min(distances))
+        else:
+            pred_index = np.argmin(distances)
+            pred_name = faces[pred_index].name
+            print('Match')
+            print('min distance:', np.min(distances))
+
+        # image_test = draw_bounding_box(image_test, loc_test)
+        # image_test = draw_name(image_test, loc_test, pred_name)
+
+    return pred_name
+
+'''
+def draw_bounding_box(image_test, loc_test):
+    top, right, bottom, left = loc_test
+
+    line_color = (0, 255, 0)
+    line_thickness = 2
+
+    cv2.rectangle(image_test, (left, top), (right, bottom), line_color, line_thickness)
+    return image_test
+
+def draw_name(image_test, loc_test, pred_name):
+    top, right, bottom, left = loc_test 
+
+    font = cv2.FONT_HERSHEY_SIMPLEX
+    font_scale = 1.5
+    font_color = (0, 0, 255)
+    line_thickness = 3
+
+    text_size, _ = cv2.getTextSize(pred_name, font, font_scale, line_thickness)
+
+    bg_top_left = (left, top-text_size[1])
+    bg_bottom_right = (left+text_size[0], top)
+    cv2.rectangle(image_test, bg_top_left, bg_bottom_right, (0, 255, 0), -1)   
+
+    cv2.putText(image_test, pred_name, (left, top), font, font_scale, font_color, line_thickness)
+
+    return image_test
+'''
+filenames = os.listdir('templates')
+images = [load_image(f'templates/{filename}') for filename in filenames]
+# show_image(images[0])
+faces = create_database(filenames, images)
+
+# My modification
+tests = os.listdir('tests')
+for test in tests:
+    if(".jpg" not in test and ".png" not in test):
+        continue
+    print(f'test: {test}')
+    image_test = load_image(f'tests/{test}')
+    # show_image(image_test)
+    name = detect_faces(image_test, faces, threshold=0.5)
+    print(f'Pridict name: {name}\n')
+
+# cap = cv2.VideoCapture(0)
+# cap.set(3, 640)
+# cap.set(4, 480)
+
+# while True:
+#     _, image, = cap.read()
+#     image = cv2.flip(image, flipCode=1)
+#     image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+
+#     image = detect_faces(image, faces, threshold=0.6)
+
+#     image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
+
+#     if cv2.waitKey(1) & 0xFF == ord('q'):
+#         break
+
+#     cv2.imshow('image', image)