HOME/Articles/

pil example expressive (snippet)

Author
natamacm
Published
2020/03/02
Updated
2020/04/06
python
Article Outline

Python pil example 'expressive'

Functions in program:

  • def capture():
  • def plot_polar(values, gender):
  • def input_JP_text(name, img, text, x, y, size, color, display):
  • def scoring():

Modules used in program:

  • import shutil
  • import os
  • import cv2
  • import collections
  • import numpy as np
  • import japanize_matplotlib
  • import matplotlib.pyplot as plt
  • import cognitive_face as CF

python expressive

Python pil example: expressive

from PIL import ImageFont, ImageDraw, Image
import cognitive_face as CF
import matplotlib.pyplot as plt
import japanize_matplotlib
import numpy as np
import collections
import cv2
import os
import shutil

KEY = '********************************'
BASE_URL = 'https://japaneast.api.cognitive.microsoft.com/face/v1.0'

CF.Key.set(KEY)
CF.BaseUrl.set(BASE_URL)

cascade_file = '/usr/local/share/OpenCV/lbpcascades/lbpcascade_frontalface.xml'
cascade = cv2.CascadeClassifier(cascade_file)

img_url = 'photo.png'

# 設定パラメータ
ESC_KEY = 27
ENTER_KEY = 13
WINDOW_NAME = "Facial muscle test"
WINDOW_WIDTH = 1024
WINDOW_HEIGHT = 768
GUIDE_WINDOW_NAME = "Guide"
GUIDE_WINDOW_WIDTH = 728
GUIDE_WINDOW_HEIGHT = 800
ANIMATION_FRAME = 40


# ベース画像の準備
base = np.zeros((GUIDE_WINDOW_HEIGHT, GUIDE_WINDOW_WIDTH, 3), dtype=np.uint8)
base.fill(255)

dark = np.zeros((WINDOW_HEIGHT, WINDOW_WIDTH, 3), dtype=np.uint8)
dark.fill(156)


def scoring():
    age = 0
    gender = []
    score = []

    # FaceAPIで取得する感情
    attributes = ['anger', 'contempt', 'disgust', 'fear', 'smile', 'happiness', 'neutral', 'sadness', 'surprise']

    # 感情を提示し、それに応じた写真を送信
    for face in attributes:
        print(face)

        faces = []
        while faces == []:
            # ガイドの画像を表示
            cv2.imshow(GUIDE_WINDOW_NAME, cv2.resize(cv2.imread('img/{}.png'.format(face)), (GUIDE_WINDOW_WIDTH, GUIDE_WINDOW_HEIGHT)))

            # カメラ起動
            capture()

            # FaceAPIに投げる
            faces = CF.face.detect(img_url, face_id=True, landmarks=False, attributes='age,gender,smile,emotion')

            # 顔を認識できなかったときはやり直す
            if faces == []:
                cv2.imshow(GUIDE_WINDOW_NAME, fail)
                cv2.waitKey(0)


        # 顔を2つ以上認識したとき
        if len(faces) >= 2:
            face_width = 0
            for kao in faces:
                if kao['faceRectangle']['width'] > face_width:
                    face_width = kao['faceRectangle']['width']
                    temp_face = kao

            # 写真内で一番大きい顔を選択
            try:
                faces = [temp_face]
                print('choice')
            except:
                print('error')
                face = [faces[0]]


        # 笑顔のときだけ別処理
        if face == 'smile':
            score.append(round(faces[0]['faceAttributes']['smile'] * 100))
            print('{}\n'.format(faces[0]['faceAttributes']['smile']))
        else:
            score.append(round(faces[0]['faceAttributes']['emotion'][face] * 100))
            print('{}\n'.format(faces[0]['faceAttributes']['emotion']))

        age += faces[0]['faceAttributes']['age']
        gender.append(faces[0]['faceAttributes']['gender'])


    # 顔写真を削除
    os.remove('photo.png')

    # 精度向上の為に9回の平均値を取る
    age = round(age // 9)

    # 精度向上の為に9回で最も多かった性別を選択する
    gender = collections.Counter(gender).most_common()[0][0]

    return score, age, gender


def input_JP_text(name, img, text, x, y, size, color, display):
    fontpath = 'font/APJapanesefont.ttf'
    font = ImageFont.truetype(fontpath, size)

    img_pil = Image.fromarray(img)
    draw = ImageDraw.Draw(img_pil)

    draw.text((x, y), text, font=font, fill=color)
    output_img = np.array(img_pil)

    if display == 1:
        cv2.imshow(name, output_img)
        cv2.waitKey(1)

    return output_img


def plot_polar(values, gender):
    input_JP_text(WINDOW_NAME, dark, '採点中', 210, 270, 200, (0, 0, 0), 1)
    cv2.imshow(GUIDE_WINDOW_NAME, progress)
    cv2.waitKey(1)

    animation = []

    labels = ['怒り', '軽蔑', '嫌気', '恐怖', '笑顔', '幸福', '真顔', '悲しみ', '驚き']
    dummy_labels = ['', '', '', '', '', '', '', '', '']

    angles = np.linspace(0, 2 * np.pi, len(labels) + 1, endpoint=True)
    values = np.concatenate((values, [values[0]]))
    fig = plt.figure(figsize=(10.24, 7.68), dpi=100)
    ax = fig.add_subplot(111, polar=True)

    ax.set_thetagrids(angles[:-1] * 180 / np.pi, dummy_labels)
    ax.set_rgrids(np.linspace(0, 100, 6), size=20, color='gray')


    # ラベルの描画
    for idx, label in enumerate(labels):
        ax.text(angles[idx], 125, label, size=30, color='black', verticalalignment='center', horizontalalignment='center')


    # アニメーション作成
    for i in range(1, ANIMATION_FRAME+1):

        # 男性は青、女性は赤
        if gender == 'male':
            ax.fill(angles, values/ANIMATION_FRAME*i, 'b', alpha=0.05)
        elif gender == 'female':
            ax.fill(angles, values/ANIMATION_FRAME*i, 'm', alpha=0.05)


        # アニメーション用の画像を保存
        fig.savefig('anime/{}.png'.format(i))


        # 進行度を描画
        input_JP_text(GUIDE_WINDOW_NAME, progress, '{}%'.format(round(i/ANIMATION_FRAME*100)), 340, 430, 40, (0, 0, 0), 1)


    # 性別に応じて外枠を描画
    if gender == 'male':
        ax.plot(angles, values, 'bo-', linewidth=4)
    elif gender == 'female':
        ax.plot(angles, values, 'mo-', linewidth=4)


    # 点数を描画
    for idx, num in enumerate(values):
        ax.text(angles[idx], num+10, num, size=20, color='black', verticalalignment='center', horizontalalignment='center')


    # 完成したレーダーチャートを保存
    fig.savefig('rader.png')


    cv2.imshow(GUIDE_WINDOW_NAME, result)
    cv2.waitKey(1)
    plt.close(fig)


def capture():
    end_flag, c_frame = cap.read()

    while end_flag == True:
        img = c_frame
        img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
        face_list = cascade.detectMultiScale(img_gray, minSize=(100, 100))


        # 顔部分を赤枠で囲む
        for (x, y, w, h) in face_list:
            color = (0, 0, 225)
            pen_w = 3
            cv2.rectangle(c_frame, (x, y), (x+w, y+h), color, thickness = pen_w)


        # そのまま描画すると左右逆で混乱するので反転する
        c_frame = cv2.resize(cv2.flip(c_frame, 1), (WINDOW_WIDTH, WINDOW_HEIGHT))
        cv2.imshow(WINDOW_NAME, c_frame)
        key = cv2.waitKey(1)


        # Escでプログラム終了
        if key == ESC_KEY:
            cv2.destroyAllWindows()
            cap.release()
            print('exit')
            exit(0)


        # エンターキーで写真撮影
        elif key == ENTER_KEY:
            cv2.imwrite(img_url, c_frame)
            break

        end_flag, c_frame = cap.read()

    # 写真撮影から抜けるときはカメラを暗くする
    blended = cv2.addWeighted(c_frame, 0.4, dark, 0.3, 0)

    cv2.imshow(WINDOW_NAME, blended)
    cv2.imshow(GUIDE_WINDOW_NAME, progress)
    cv2.waitKey(1)



if __name__ == '__main__':
    # カメラとガイドの準備
    cap = cv2.VideoCapture(0)
    cv2.namedWindow(WINDOW_NAME)
    cv2.moveWindow(WINDOW_NAME, 0, 0)
    cv2.namedWindow(GUIDE_WINDOW_NAME)
    cv2.moveWindow(GUIDE_WINDOW_NAME, 1024, 0)


    # レーダーチャート検証用
    gender = 'male'
    score = [85, 78, 66, 100, 88, 64, 73, 82, 91]


    # ガイド用画像の準備
    progress = input_JP_text(GUIDE_WINDOW_NAME, base, '少々お待ちください。', 110, 330, 70, (0, 0, 0), 0)
    start = input_JP_text(GUIDE_WINDOW_NAME, base, '表情筋テスト', 110, 300, 100, (0, 0, 0), 0)
    start = input_JP_text(GUIDE_WINDOW_NAME, start, 'ボタンを押してください', 220, 420, 35, (0, 0, 0), 0)
    result = input_JP_text(GUIDE_WINDOW_NAME, base, '結果が出ました！', 150, 330, 70, (0, 0, 0), 0)
    result = input_JP_text(GUIDE_WINDOW_NAME, result, 'ボタンを押すと終了します', 220, 420, 30, (0, 0, 0), 0)
    standby = input_JP_text(WINDOW_NAME, dark, '待機中', 210, 250, 200, (0, 0, 0), 0)
    fail = input_JP_text(GUIDE_WINDOW_NAME, base, '顔を認識できませんでした', 130, 330, 50, (0, 0, 0), 0)
    fail = input_JP_text(GUIDE_WINDOW_NAME, fail, 'ボタンを押してください', 250, 420, 30, (0, 0, 0), 0)


    while True:
        # スタート画面
        cv2.imshow(GUIDE_WINDOW_NAME, start)
        input_JP_text(WINDOW_NAME, standby, 'ボタンを押してください', 290, 470, 50, (0, 0, 0), 1)

        key = cv2.waitKey(0)

        # Escでプログラム終了
        if key == ESC_KEY:
            break


        # 表情筋テスト開始
        score, age, gender = scoring()        


        # レーダーチャート作成
        plot_polar(score, gender)

        for i in range(ANIMATION_FRAME):
            cv2.imshow(WINDOW_NAME, cv2.imread('anime/{}.png'.format(i+1)))
            cv2.waitKey(1)

        cv2.imshow(WINDOW_NAME, cv2.imread('rader.png'))
        cv2.waitKey(0)


        # レーダーチャートを連番で保存
        files = os.listdir('rader')
        num = int(sorted(files, reverse=True)[0].replace('.png', ''))+1
        shutil.copy('rader.png', 'rader/{:04}.png'.format(num))


    # 終了処理
    cv2.destroyAllWindows()
    cap.release()
BACK TO TOP
© natamacm 2019
Powered by GitPress.io
Sponsored by Quail.ink