# import the necessary packages
import numpy as np
import argparse
import time
import cv2
import os

image_path='11.jpg'
yolo='yolo_coco'
confidence_t=0.5
threshold=0.3
# 加載訓(xùn)練 YOLO 模型的 COCO 類標簽
labelsPath = os.path.sep.join([yolo, "coco.names"])
LABELS = open(labelsPath).read().strip().split("\n")
# 初始化一個顏色列表來表示每個類標簽
np.random.seed(42)
COLORS = np.random.randint(0, 255, size=(len(LABELS), 3),
                           dtype="uint8")
# YOLO 對象檢測
print("[INFO] loading YOLO from disk...")
config_path = './yolo_coco/yolov3.cfg'
weights_path = './yolo_coco/yolov3.weights'
net = cv2.dnn.readNetFromDarknet(config_path, weights_path)

# 加載我們的輸入圖像并獲取其空間維度
image = cv2.imread(image_path)
(H, W) = image.shape[:2]
# 從輸入圖像構(gòu)建一個blob，然后執(zhí)行一個前向傳播
# 通過 YOLO 對象檢測器，輸出邊界框和相關(guān)概率
blob = cv2.dnn.blobFromImage(image, 1 / 255.0, (416, 416),
                             swapRB=True, crop=False)
net.setInput(blob)
start = time.time()
# 獲取網(wǎng)絡(luò)輸出層信息（所有輸出層的名字），設(shè)定并前向傳播
outInfo = net.getUnconnectedOutLayersNames()
# 得到各個輸出層的、各個檢測框等信息，是二維結(jié)構(gòu)。
layerOutputs = net.forward(outInfo)

# 分別初始化檢測到的邊界框、置信度和類 ID 的列表
boxes = []
confidences = []
classIDs = []
# 循環(huán)輸出
for output in layerOutputs:
    # 遍歷每個檢測結(jié)果
    for detection in output:
        # 提取物體檢測的類ID和置信度（即概率）
        scores = detection[5:]
        classID = np.argmax(scores)
        confidence = scores[classID]
        # 過濾精度低的結(jié)果
        if confidence > confidence_t:
            # 延展邊界框坐標，計算 YOLO 邊界框的中心 (x, y) 坐標，然后是框的寬度和高度
            box = detection[0:4] * np.array([W, H, W, H])
            (centerX, centerY, width, height) = box.astype("int")
            # 使用中心 (x, y) 坐標導(dǎo)出邊界框的上角和左角
            x = int(centerX - (width / 2))
            y = int(centerY - (height / 2))
            # 更新邊界框坐標、置信度和類 ID 列表
            boxes.append([x, y, int(width), int(height)])
            confidences.append(float(confidence))
            classIDs.append(classID)
# 使用非極大值抑制來抑制弱的、重疊的邊界框
idxs = cv2.dnn.NMSBoxes(boxes, confidences, confidence_t,
                        threshold)
# 確保至少存在一個檢測
if len(idxs) > 0:
    # 遍歷我們保存的索引
    for i in idxs.flatten():
        # 提取邊界框坐標
        (x, y) = (boxes[i][0], boxes[i][1])
        (w, h) = (boxes[i][2], boxes[i][3])
        # 在圖像上繪制一個邊界框矩形和標簽
        color = [int(c) for c in COLORS[classIDs[i]]]
        cv2.rectangle(image, (x, y), (x + w, y + h), color, 2)
        text = "{}: {:.4f}".format(LABELS[classIDs[i]], confidences[i])
        cv2.putText(image, text, (x, y - 5), cv2.FONT_HERSHEY_SIMPLEX,
                    0.5, color, 2)
# show the output image
cv2.imshow("Image", image)
cv2.waitKey(0)

import numpy as np
import imutils
import time
import cv2
import os

yolo = 'yolo_coco'
confidence_t = 0.5
threshold = 0.3
output = 'output.avi'

# 加載YOLO 模型訓(xùn)練的 COCO 類標簽
labelsPath = os.path.sep.join([yolo, "coco.names"])
LABELS = open(labelsPath).read().strip().split("\n")
# 初始化顏色列表
np.random.seed(42)
COLORS = np.random.randint(0, 255, size=(len(LABELS), 3), dtype="uint8")
# 配置 YOLO 權(quán)重和模型配置的路徑
weightsPath = os.path.sep.join([yolo, "yolov3.weights"])
configPath = os.path.sep.join([yolo, "yolov3.cfg"])
# 加載在 COCO 數(shù)據(jù)集（80 個類）上訓(xùn)練的 YOLO 對象檢測，并獲取YOLO輸出層的名稱
net = cv2.dnn.readNetFromDarknet(configPath, weightsPath)
# 獲取網(wǎng)絡(luò)輸出層信息（所有輸出層的名字），設(shè)定并前向傳播
outInfo = net.getUnconnectedOutLayersNames()
# 初始化視頻流、指向輸出視頻文件的指針和幀尺寸
vs = cv2.VideoCapture(0)
writer = None
(W, H) = (None, None)
# 獲取文件的總幀數(shù)。
try:
    prop = cv2.cv.CV_CAP_PROP_FRAME_COUNT if imutils.is_cv2() \
        else cv2.CAP_PROP_FRAME_COUNT
    total = int(vs.get(prop))
    print("[INFO] {} total frames in video".format(total))
except:
    print("[INFO] could not determine # of frames in video")
    print("[INFO] no approx. completion time can be provided")
    total = -1

# loop over frames from the video file stream
while True:
    # 從文件中讀取下一幀
    (grabbed, frame) = vs.read()
    # 如果幀沒有被抓取，那么已經(jīng)到了流的末尾
    if not grabbed:
        break
    # 如果框架尺寸為空，則給他們賦值
    if W is None or H is None:
        (H, W) = frame.shape[:2]
    # 從輸入幀構(gòu)造一個 blob，然后執(zhí)行 YOLO 對象檢測器的前向傳遞，得到邊界框和相關(guān)概率
    blob = cv2.dnn.blobFromImage(frame, 1 / 255.0, (416, 416),
                                 swapRB=True, crop=False)
    net.setInput(blob)
    start = time.time()
    layerOutputs = net.forward(outInfo)
    end = time.time()
   # 分別初始化檢測到的邊界框、置信度和類 ID 的列表
    boxes = []
    confidences = []
    classIDs = []
    # 循環(huán)輸出
    for output in layerOutputs:
        # 遍歷每個檢測結(jié)果
        for detection in output:
            # 提取物體檢測的類ID和置信度（即概率）
            scores = detection[5:]
            classID = np.argmax(scores)
            confidence = scores[classID]
             # 過濾精度低的結(jié)果
            if confidence > confidence_t:
               # 縮放邊界框坐標，計算 YOLO 邊界框的中心 (x, y) 坐標，然后是框的寬度和高度
                box = detection[0:4] * np.array([W, H, W, H])
                (centerX, centerY, width, height) = box.astype("int")
                # 使用中心 (x, y) 坐標導(dǎo)出邊界框的上角和左角
                x = int(centerX - (width / 2))
                y = int(centerY - (height / 2))
               # 更新邊界框坐標、置信度和類 ID 列表
                boxes.append([x, y, int(width), int(height)])
                confidences.append(float(confidence))
                classIDs.append(classID)
    # 使用非極大值抑制來抑制弱的、重疊的邊界框
    idxs = cv2.dnn.NMSBoxes(boxes, confidences, confidence_t,
                            threshold)
    # 確保至少存在一個檢測
    if len(idxs) > 0:
        # 遍歷保存的索引
        for i in idxs.flatten():
            # 在圖像上繪制一個邊界框矩形和標簽
            (x, y) = (boxes[i][0], boxes[i][1])
            (w, h) = (boxes[i][2], boxes[i][3])
           # 確保至少存在一個檢測
            color = [int(c) for c in COLORS[classIDs[i]]]
            cv2.rectangle(frame, (x, y), (x + w, y + h), color, 2)
            text = "{}: {:.4f}".format(LABELS[classIDs[i]],
                                       confidences[i])
            cv2.putText(frame, text, (x, y - 5),
                        cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2)
    cv2.imshow("Frame", frame)
    key = cv2.waitKey(1) & 0xFF
    # check if the video writer is None
    if writer is None:
        # initialize our video writer
        fourcc = cv2.VideoWriter_fourcc(*'XVID')
        writer = cv2.VideoWriter('output.avi', fourcc, 30, (int(frame.shape[1]), int(frame.shape[0])))
        # some information on processing single frame
        if total > 0:
            elap = (end - start)
            print("[INFO] single frame took {:.4f} seconds".format(elap))
            print("[INFO] estimated total time to finish: {:.4f}".format(
                elap * total))
    # write the output frame to disk
    writer.write(frame)
# release the file pointers
print("[INFO] cleaning up...")
writer.release()
vs.release()