文件編號：A343

文件大?。?/strong>149M

開發(fā)環(huán)境：Python3.7、OpenCV4.5、dlib、Pycharm2020

猿創(chuàng)承諾：該項目親測正常運行，需遠程調(diào)試部署需另外收費，確保正常使用，不能正常使用全額退款。

簡要概述：利用 Python 開發(fā)，借助 Dlib 庫捕獲攝像頭中的人臉，提取人臉特征，通過計算特征值之間的歐氏距離，來和預(yù)存的人臉特征進行對比，判斷是否匹配，達到人臉識別的目的；

- 完成的功能

    - 新的人臉數(shù)據(jù)集和識別模型

    - 做到視頻識別， 實時顯示人臉信息

    - 可攝像頭添加人臉， 且可改名

    - 可通過文件夾方式批量添加人臉

    - GUI化， 基本無卡頓

.. image:: introduction/main_interface.png

  :align: center

.. image:: introduction/main_figure_interface.png

  :align: center

關(guān)于算法：

- 基于 Residual Neural Network / 殘差網(wǎng)絡(luò)的 CNN 模型;

- This model is a ResNet network with 29 conv layers. It's essentially a version of the ResNet-34 network from the paper Deep Residual Learning for Image Recognition by He, Zhang, Ren, and Sun with a few layers removed and the number of filters per layer reduced by half.

實現(xiàn)流程：

- 安裝依賴庫

- 進行人臉信息采集錄入

- 提取所有錄入人臉數(shù)據(jù)存入 "features_all.csv"

- 調(diào)用攝像頭進行實時人臉識別

關(guān)于代碼：

#. 樹狀圖:

    ├── main_interface.py               # Step0. Start First

    ├── get_faces_from_camera.py        # Step1. Faces register

    ├── features_extraction_to_csv.py   # Step2. Features extraction

    ├── face_reco_from_camera.py        # Step3. Faces recognition

    ├── how_to_use_camera.py            # Use the default camera by opencv

    ├── data

    │   ├── data_dlib                   # Dlib's model

    │   │   ├── dlib_face_recognition_resnet_model_v1.dat

    │   │   ├── shape_predictor_5_face_landmarks.dat

    │   │   └── shape_predictor_68_face_landmarks.dat

    │   ├── data_faces_from_camera      # Face images captured from camera (will generate after step 1)

    │   │   ├── person_1

    │   │   │   ├── img_face_1.jpg

    │   │   │   └── img_face_2.jpg

    │   │   └── person_2

    │   │       └── img_face_1.jpg

    │   │       └── img_face_2.jpg

    │   └── features_all.csv            # CSV to save all the features of known faces (will generate after step 2)

    ├── introduction                    # Some files for readme.rst

    │   ├── main_figure_interface.png

    │   ├── main_interface.png

    │   └── overview.png

    └── requirements.txt                # Some python packages needed

用到的 Dlib 相關(guān)模型函數(shù):

#. Dlib 正向人臉檢測器 (based on HOG), output: <class 'dlib.dlib.rectangles'>

   .. code-block:: python

      detector = dlib.get_frontal_face_detector()

      faces = detector(img_gray, 0)

#. Dlib 人臉預(yù)測器, output: <class 'dlib.dlib.full_object_detection'>,

   will use shape_predictor_68_face_landmarks.dat

   .. code-block:: python

      # This is trained on the ibug 300-W dataset (https://ibug.doc.ic.ac.uk/resources/facial-point-annotations/)

      # Also note that this model file is designed for use with dlib's HOG face detector.

      # That is, it expects the bounding boxes from the face detector to be aligned a certain way, the way dlib's HOG face detector does it.

      # It won't work as well when used with a face detector that produces differently aligned boxes,

      # such as the CNN based mmod_human_face_detector.dat face detector.

      predictor = dlib.shape_predictor("data/data_dlib/shape_predictor_68_face_landmarks.dat")

      shape = predictor(img_rd, faces[i])

#. 特征描述子 Face recognition model, the object maps human faces into 128D vectors

   .. code-block:: python

      face_rec = dlib.face_recognition_model_v1("data/data_dlib/dlib_face_recognition_resnet_model_v1.dat")

源碼介紹:

#. get_face_from_camera.py: 

   進行 Face register / 人臉信息采集錄入

   * 請注意存儲人臉圖片時，矩形框不要超出攝像頭范圍，要不然無法保存到本地;

   * 超出會有 "out of range" 的提醒;

#. features_extraction_to_csv.py:

   從上一步存下來的圖像文件中，提取人臉數(shù)據(jù)存入CSV;

   * 會生成一個存儲所有特征人臉數(shù)據(jù)的 "features_all.csv"；

   * size: n*128 , n means n people you registered and 128 means 128D features of the face

#. face_reco_from_camera.py: 

   這一步將調(diào)用攝像頭進行實時人臉識別; / This part will implement real-time face recognition;

   * Compare the faces captured from camera with the faces you have registered which are saved in "features_all.csv"

   * 將捕獲到的人臉數(shù)據(jù)和之前存的人臉數(shù)據(jù)進行對比計算歐式距離, 由此判斷是否是同一個人;

按需寫作：

演示視頻：

點擊查看：系統(tǒng)演示視頻

運行效果：

遠程協(xié)助：

溫馨提示：索煒達.猿創(chuàng)官方提供收費遠程協(xié)助，確保您項目運行成功。

點擊查看：遠程協(xié)助相關(guān)事項

我們提供完整項目文件清單如下：

文件目錄

 ├ 1.項目源碼

 ├ 2.運行截圖

 └ 3.演示視頻