Remove face_recognition module

.gitignore

@@ -99,3 +99,9 @@ ENV/
 
 # mypy
 .mypy_cache/
+
+#pycharm config files
+/.idea
+.idea/
+#train images
+images/

faceai/FaceRecognition/FaceRecognition.py

@@ -0,0 +1,94 @@
+import dlib
+import sys
+sys.path.append("..")
+from tool import compare, imtool, save_load
+
+
+class FaceRecognition:
+    def __init__(self, shape_predict_path, recognition_model_path):
+        self.shape_detector = dlib.get_frontal_face_detector()
+        self.shape_predictor = dlib.shape_predictor(shape_predict_path)
+        self.face_recognition = dlib.face_recognition_model_v1(recognition_model_path)
+        self.train_data = []
+        self.train_label = []
+
+    def train(self, train_data, is_debug=False):
+        if not isinstance(train_data, list):
+            raise TypeError("training_data required to be a list which contains labels and images")
+
+        self.train_data = []
+        self.train_label = []
+        for data in train_data:
+            if is_debug:
+                print("Processing {}".format(data.get("label")))
+            for img in data.get("data"):
+                encodings, bounds = self._get_face_encoding(img)
+                self.train_data = self.train_data + encodings
+                self.train_label = self.train_label + [data.get("label") for i in encodings]
+
+        return self
+
+    def _get_face_encoding(self, image):
+        bounds = []
+        encodings = []
+        detect_faces = self.shape_detector(image, 1)
+        for face in detect_faces:
+            shape = self.shape_predictor(image, face)
+            encodings.append(self.face_recognition.compute_face_descriptor(image, shape))
+            bounds.append((face.left(), face.top(), face.right(), face.bottom()))
+
+        return encodings, bounds
+
+    def predict(self, image, threshold=0.6):
+        encodings, bounds = self._get_face_encoding(image)
+        predict_info = []
+        for i, vector in enumerate(encodings):
+            sort_indexes, distances = compare.compare(self.train_data, vector)
+            distance = distances[sort_indexes[0]]
+            if distance < threshold:
+                predict_info.append((self.train_label[sort_indexes[0]], bounds[i]))
+            else:
+                predict_info.append((None, bounds[i]))
+
+        return predict_info
+
+    def save(self, file_name):
+        data = {
+            "train_data": self.train_data,
+            "train_label": self.train_label
+        }
+        save_load.save(data, file_name)
+        return self
+
+    def load(self, file_name):
+        data = save_load.load(file_name)
+        self.train_data = data.get("train_data")
+        self.train_label = data.get("train_label")
+        return self
+
+
+if __name__ == '__main__':
+    from tool import file
+    import argparse
+
+    parse = argparse.ArgumentParser()
+    parse.add_argument("dlib_face_path", help="Please provide the dlib_face_recognition_model path, you can download it from http://dlib.net/files/dlib_face_recognition_resnet_model_v1.dat.bz2 and extract it.")
+    parse.add_argument("dlib_shape_path", help="Please provide the dlib_shape_detector_model path, you can download it from http://dlib.net/files/shape_predictor_5_face_landmarks.dat.bz2 and extract it.")
+    parse.add_argument("train_path", help="Please provide the training images folder path.")
+    parse.add_argument("test_path", help="Please provide the testing images folder path.")
+    args = parse.parse_args()
+
+    recognition_path = args.dlib_face_path #"../dlib_model/dlib_face_recognition_resnet_model_v1.dat"
+    shape_path = args.dlib_shape_path #"../dlib_model/shape_predictor_5_face_landmarks.dat"
+    train_data = file.get_images_data(args.train_path) #("../../images/train")
+    test_data = file.get_images_data(args.test_path) #("../../images/test")
+
+    face = FaceRecognition(shape_path, recognition_path)
+    face.train(train_data, is_debug=True)
+
+    for data in test_data:
+        for image in data.get("data"):
+            for label, bounds in face.predict(image):
+                imtool.put_text(image, label if label is not None else "Unknown", bounds)
+                imtool.draw_rect(image, bounds)
+                imtool.show_image(image, width=400)

faceai/tool/__init__.py

@@ -0,0 +1,4 @@
+#
+# coding=utf-8
+# description: tool package init file
+#

faceai/tool/compare.py

@@ -0,0 +1,26 @@
+#
+# coding=utf-8
+# description: calculate and compare the distance of multiple 2-d vectors.
+#
+
+import numpy as np
+
+
+def distance(vector1, vector2):
+    if not isinstance(vector1, np.ndarray):
+        vector1 = np.array(vector1)
+    if not isinstance(vector2, np.ndarray):
+        vector2 = np.array(vector2)
+
+    return np.linalg.norm(vector1 - vector2)
+
+
+def compare(vectors, vector):
+    distances = []
+    for v in vectors:
+        distances.append(distance(v, vector))
+    return np.argsort(distances), distances
+
+
+if __name__ == '__main__':
+    pass

faceai/tool/file.py

@@ -0,0 +1,45 @@
+#
+# coding=utf-8
+# description: file operation functions
+#
+
+import os
+import cv2
+
+
+IMAGE_SUFFIX = ["jpg", "jpeg", "png"]
+
+
+def is_image(image_name):
+    return any([image_name.lower().endswith(suffix) for suffix in IMAGE_SUFFIX])
+
+
+def get_images_data(train_path, image_flag=cv2.IMREAD_COLOR):
+    """
+    loading training images or test images.
+    :param train_path:
+    :param image_flag:
+    :return:
+    """
+    if not os.path.isdir(train_path):
+        raise NotADirectoryError(train_path + " is not a dir path.")
+
+    data = []
+
+    for path in os.listdir(train_path):
+        name = path
+        path = os.path.join(train_path, path)
+        images = []
+        data.append({
+            "label": name,
+            "data": images
+        })
+        for image_name in os.listdir(path):
+            if not is_image(image_name):
+                continue
+            images.append(cv2.imread(os.path.join(path, image_name), image_flag))
+    return data
+
+
+if __name__ == '__main__':
+    pass

faceai/tool/imtool.py

@@ -0,0 +1,39 @@
+#
+# coding=utf-8
+# description:
+#
+import cv2
+
+
+def show_image(img, width=None, height=None, win_name="Image", wait_time=0):
+    cv2.imshow(win_name, resize(img, width, height))
+    cv2.waitKey(wait_time) & 0XFF
+
+
+def resize(img, width, height, interpolation=cv2.INTER_AREA):
+    size = None
+
+    if width is None and height is None:
+        size = (img.shape[1], img.shape[0])
+    elif width is not None:
+        ratio = width / float(img.shape[1])
+        height = int(img.shape[0] * ratio)
+        size = (width, height)
+    elif height is not None:
+        ratio = height / float(img.shape[0])
+        width = int(img.shape[1] * ratio)
+        size = (width, height)
+
+    return cv2.resize(img, size, interpolation=interpolation)
+
+
+def draw_rect(img, rect, color=(0, 255, 0), line_width=1):
+    cv2.rectangle(img, rect[:2], rect[2:], color, line_width)
+
+
+def put_text(img, text, rect, color=(0, 255, 0), thickness=0.8):
+    cv2.putText(img, text, rect[:2], cv2.FONT_HERSHEY_COMPLEX_SMALL, thickness, color)
+
+
+if __name__ == '__main__':
+    pass

faceai/tool/save_load.py

@@ -0,0 +1,20 @@
+#
+# coding=utf-8
+# description:
+#
+import pickle
+
+
+def load(file_name):
+    with open(file_name, "rb") as fs:
+        data = pickle.load(fs)
+    return data
+
+
+def save(data, file_name):
+    with open(file_name, "wb") as fs:
+        pickle.dump(data, fs)
+
+
+if __name__ == '__main__':
+    pass