目標檢測 YOLO v3 驗證 COCO 模型

SpikeKing發表於2019-03-04

YOLO,是You Only Look Once的縮寫,一種基於深度卷積神經網路的物體檢測演算法,YOLO v3是YOLO的第3個版本,檢測演算法更快更準,2018年4月8日。

本文原始碼:https://github.com/SpikeKing/keras-yolo3-detection

歡迎Follow我的GitHub:https://github.com/SpikeKing

資料集

YOLO v3已經提供COCO(Common Objects in Context)資料集的模型引數,支援直接用於物體檢測,模型248M,下載:

wget https://pjreddie.com/media/files/yolov3.weights
複製程式碼

將模型引數轉換為Keras的模型引數,模型248.6M,轉換:

python convert.py -w yolov3.cfg model_data/yolov3.weights model_data/yolo_weights.h5
複製程式碼

畫出網路結構:

plot_model(model, to_file=`./model_data/model.png`, show_shapes=True, show_layer_names=True)  # 網路圖
複製程式碼

COCO含有80個類別:

person(人)  

bicycle(自行車)  car(汽車)  motorbike(摩托車)  aeroplane(飛機)  bus(公共汽車)  train(火車)  truck(卡車)  boat(船)  

traffic light(訊號燈)  fire hydrant(消防栓)  stop sign(停車標誌)  parking meter(停車計費器)  bench(長凳)  

bird(鳥)  cat(貓)  dog(狗)  horse(馬)  sheep(羊)  cow(牛)  elephant(大象)  bear(熊)  zebra(斑馬)  giraffe(長頸鹿)  

backpack(揹包)  umbrella(雨傘)  handbag(手提包)  tie(領帶)  suitcase(手提箱)  

frisbee(飛盤)  skis(滑雪板雙腳)  snowboard(滑雪板)  sports ball(運動球)  kite(風箏) baseball bat(棒球棒)  baseball glove(棒球手套)  skateboard(滑板)  surfboard(衝浪板)  tennis racket(網球拍)  

bottle(瓶子)  wine glass(高腳杯)  cup(茶杯)  fork(叉子)  knife(刀)
spoon(勺子)  bowl(碗)  

banana(香蕉)  apple(蘋果)  sandwich(三明治)  orange(橘子)  broccoli(西蘭花)  carrot(胡蘿蔔)  hot dog(熱狗)  pizza(披薩)  donut(甜甜圈)  cake(蛋糕)

chair(椅子)  sofa(沙發)  pottedplant(盆栽植物)  bed(床)  diningtable(餐桌)  toilet(廁所)  tvmonitor(電視機)  

laptop(筆記本)  mouse(滑鼠)  remote(遙控器)  keyboard(鍵盤)  cell phone(電話)  

microwave(微波爐)  oven(烤箱)  toaster(烤麵包器)  sink(水槽)  refrigerator(冰箱)

book(書)  clock(鬧鐘)  vase(花瓶)  scissors(剪刀)  teddy bear(泰迪熊)  hair drier(吹風機)  toothbrush(牙刷)
複製程式碼

YOLO的預設anchors是9個,對應三個尺度,每個尺度含有3個anchors,如下:

10,13,  16,30,  33,23,  30,61,  62,45,  59,119,  116,90,  156,198,  373,326
複製程式碼

檢測器

YOLO檢測類的構造器:

  1. anchors、model、classes是引數檔案,其中,anchors可以使用預設,但是model與classes必須相互匹配;
  2. score和iou是檢測引數,即置信度閾值和交叉區域閾值,置信度閾值避免誤檢,交叉區域閾值避免物體重疊;
  3. self.class_namesself.anchors,讀取類別和anchors;
  4. self.sess是TensorFlow的上下文環境;
  5. self.model_image_size,檢測圖片尺寸,將原圖片同比例resize檢測尺寸,空白填充;
  6. self.generate()是引數流程,輸出框(boxes)、置信度(scores)和類別(classes);

原始碼:

class YOLO(object):
    def __init__(self):
        self.anchors_path = `configs/yolo_anchors.txt`  # anchors
        self.model_path = `model_data/yolo_weights.h5`  # 模型檔案
        self.classes_path = `configs/coco_classes.txt`  # 類別檔案

        self.score = 0.3  # 置信度閾值
        # self.iou = 0.45
        self.iou = 0.20  # 交叉區域閾值

        self.class_names = self._get_class()  # 獲取類別
        self.anchors = self._get_anchors()  # 獲取anchor
        self.sess = K.get_session()
        self.model_image_size = (416, 416)  # fixed size or (None, None), hw
        self.boxes, self.scores, self.classes = self.generate()

    def _get_class(self):
        classes_path = os.path.expanduser(self.classes_path)
        with open(classes_path) as f:
            class_names = f.readlines()
        class_names = [c.strip() for c in class_names]
        return class_names

    def _get_anchors(self):
        anchors_path = os.path.expanduser(self.anchors_path)
        with open(anchors_path) as f:
            anchors = f.readline()
        anchors = [float(x) for x in anchors.split(`,`)]
        return np.array(anchors).reshape(-1, 2)
複製程式碼

引數流程:輸出框(boxes)、置信度(scores)和類別(classes)

  1. yolo_body網路中,載入yolo_model引數;
  2. 為不同的框,生成不同的顏色,隨機;
  3. 將模型的輸出,經過置信度和交叉區域,過濾框;

原始碼:

def generate(self):
    model_path = os.path.expanduser(self.model_path)  # 轉換~
    assert model_path.endswith(`.h5`), `Keras model or weights must be a .h5 file.`

    num_anchors = len(self.anchors)  # anchors的數量
    num_classes = len(self.class_names)  # 類別數

    # 載入模型引數
    self.yolo_model = yolo_body(Input(shape=(None, None, 3)), 3, num_classes)
    self.yolo_model.load_weights(model_path)

    print(`{} model, {} anchors, and {} classes loaded.`.format(model_path, num_anchors, num_classes))

    # 不同的框,不同的顏色
    hsv_tuples = [(float(x) / len(self.class_names), 1., 1.)
                  for x in range(len(self.class_names))]  # 不同顏色
    self.colors = list(map(lambda x: colorsys.hsv_to_rgb(*x), hsv_tuples))
    self.colors = list(map(lambda x: (int(x[0] * 255), int(x[1] * 255), int(x[2] * 255)), self.colors))  # RGB
    np.random.seed(10101)
    np.random.shuffle(self.colors)
    np.random.seed(None)

    # 根據檢測引數,過濾框
    self.input_image_shape = K.placeholder(shape=(2,))
    boxes, scores, classes = yolo_eval(self.yolo_model.output, self.anchors, len(self.class_names),
                                       self.input_image_shape, score_threshold=self.score, iou_threshold=self.iou)
    return boxes, scores, classes
複製程式碼

檢測方法detect_image

第1步,影像處理:

  1. 將影像等比例轉換為檢測尺寸,檢測尺寸需要是32的倍數,周圍進行填充;
  2. 將圖片增加1維,符合輸入引數格式;
if self.model_image_size != (None, None):  # 416x416, 416=32*13,必須為32的倍數,最小尺度是除以32
    assert self.model_image_size[0] % 32 == 0, `Multiples of 32 required`
    assert self.model_image_size[1] % 32 == 0, `Multiples of 32 required`
    boxed_image = letterbox_image(image, tuple(reversed(self.model_image_size)))  # 填充影像
else:
    new_image_size = (image.width - (image.width % 32), image.height - (image.height % 32))
    boxed_image = letterbox_image(image, new_image_size)
image_data = np.array(boxed_image, dtype=`float32`)
print(`detector size {}`.format(image_data.shape))
image_data /= 255.  # 轉換0~1
image_data = np.expand_dims(image_data, 0)  # 新增批次維度,將圖片增加1維
複製程式碼

第2步,feed資料,影像,影像尺寸;

out_boxes, out_scores, out_classes = self.sess.run(
    [self.boxes, self.scores, self.classes],
    feed_dict={
        self.yolo_model.input: image_data,
        self.input_image_shape: [image.size[1], image.size[0]],
        K.learning_phase(): 0
    })
複製程式碼

第3步,繪製邊框,自動設定邊框寬度,繪製邊框和類別文字,使用Pillow。

font = ImageFont.truetype(font=`font/FiraMono-Medium.otf`,
                          size=np.floor(3e-2 * image.size[1] + 0.5).astype(`int32`))  # 字型
thickness = (image.size[0] + image.size[1]) // 512  # 厚度
for i, c in reversed(list(enumerate(out_classes))):
    predicted_class = self.class_names[c]  # 類別
    box = out_boxes[i]  # 框
    score = out_scores[i]  # 執行度

    label = `{} {:.2f}`.format(predicted_class, score)  # 標籤
    draw = ImageDraw.Draw(image)  # 畫圖
    label_size = draw.textsize(label, font)  # 標籤文字

    top, left, bottom, right = box
    top = max(0, np.floor(top + 0.5).astype(`int32`))
    left = max(0, np.floor(left + 0.5).astype(`int32`))
    bottom = min(image.size[1], np.floor(bottom + 0.5).astype(`int32`))
    right = min(image.size[0], np.floor(right + 0.5).astype(`int32`))
    print(label, (left, top), (right, bottom))  # 邊框

    if top - label_size[1] >= 0:  # 標籤文字
        text_origin = np.array([left, top - label_size[1]])
    else:
        text_origin = np.array([left, top + 1])

    # My kingdom for a good redistributable image drawing library.
    for i in range(thickness):  # 畫框
        draw.rectangle(
            [left + i, top + i, right - i, bottom - i],
            outline=self.colors[c])
    draw.rectangle(  # 文字背景
        [tuple(text_origin), tuple(text_origin + label_size)],
        fill=self.colors[c])
    draw.text(text_origin, label, fill=(0, 0, 0), font=font)  # 文案
    del draw
複製程式碼

目標檢測

使用YOLO檢測器,檢測影像:

def detect_img_for_test(yolo):
    img_path = `./dataset/a4386X6Te9ajq866zgOtWKLx18XGW.jpg`
    image = Image.open(img_path)
    r_image = yolo.detect_image(image)
    r_image.show()
    yolo.close_session()
    
if __name__ == `__main__`:
    detect_img_for_test(YOLO())
複製程式碼

效果:

output

參考1參考2參考3,Thx@qqwweee

OK, that‘s all! Enjoy it!

相關文章