# Copyright (c) 2022 guluC
 
#導入需要的庫
import os
import sys
from pathlib import Path
import numpy as np
import cv2
import torch
import torch.backends.cudnn as cudnn

#初始化目錄
FILE = Path(__file__).resolve()
ROOT = FILE.parents[0]  # 定義YOLOv5的根目錄
if str(ROOT) not in sys.path:
    sys.path.append(str(ROOT))  # 將YOLOv5的根目錄添加到環(huán)境變量中（程序結(jié)束后刪除）
ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative
 
from models.common import DetectMultiBackend
from utils.datasets import IMG_FORMATS, VID_FORMATS, LoadImages, LoadStreams
from utils.general import (LOGGER, check_file, check_img_size, check_imshow, check_requirements, colorstr,
                           increment_path, non_max_suppression, print_args, scale_coords, strip_optimizer, xyxy2xywh)
from utils.plots import Annotator, colors, save_one_box
from utils.torch_utils import select_device, time_sync
 
#導入letterbox
from utils.augmentations import Albumentations, augment_hsv, copy_paste, letterbox, mixup, random_perspective
 
weights=ROOT / 'yolov5s.pt'  # 權重文件地址   .pt文件
source=ROOT / 'data/images'  # 測試數(shù)據(jù)文件(圖片或視頻)的保存路徑
data=ROOT / 'data/coco128.yaml'  # 標簽文件地址   .yaml文件
 
imgsz=(640, 640)  # 輸入圖片的大小 默認640(pixels)
conf_thres=0.25  # object置信度閾值 默認0.25  用在nms中
iou_thres=0.45  # 做nms的iou閾值 默認0.45   用在nms中
max_det=1000  # 每張圖片最多的目標數(shù)量  用在nms中
device='0'  # 設置代碼執(zhí)行的設備 cuda device, i.e. 0 or 0,1,2,3 or cpu
classes=None  # 在nms中是否是只保留某些特定的類 默認是None 就是所有類只要滿足條件都可以保留 --class 0, or --class 0 2 3
agnostic_nms=False  # 進行nms是否也除去不同類別之間的框 默認False
augment=False  # 預測是否也要采用數(shù)據(jù)增強 TTA 默認False
visualize=False  # 特征圖可視化 默認FALSE
half=False  # 是否使用半精度 Float16 推理 可以縮短推理時間 但是默認是False
dnn=False  # 使用OpenCV DNN進行ONNX推理
 
# 獲取設備
device = select_device(device)
 
# 載入模型
model = DetectMultiBackend(weights, device=device, dnn=dnn, data=data)
stride, names, pt, jit, onnx, engine = model.stride, model.names, model.pt, model.jit, model.onnx, model.engine
imgsz = check_img_size(imgsz, s=stride)  # 檢查圖片尺寸
 
# Half
# 使用半精度 Float16 推理
half &= (pt or jit or onnx or engine) and device.type != 'cpu'  # FP16 supported on limited backends with CUDA
if pt or jit:
    model.model.half() if half else model.model.float()
def detect(img):
    # Dataloader
    # 載入數(shù)據(jù)
    dataset = LoadImages(source, img_size=imgsz, stride=stride, auto=pt)
    # Run inference
    # 開始預測
    model.warmup(imgsz=(1, 3, *imgsz), half=half)  # warmup
    dt, seen = [0.0, 0.0, 0.0], 0
    #對圖片進行處理
    im0 = img
    # Padded resize
    im = letterbox(im0, imgsz, stride, auto=pt)[0]
    # Convert
    im = im.transpose((2, 0, 1))[::-1]  # HWC to CHW, BGR to RGB
    im = np.ascontiguousarray(im)
    t1 = time_sync()
    im = torch.from_numpy(im).to(device)
    im = im.half() if half else im.float()  # uint8 to fp16/32
    im /= 255  # 0 - 255 to 0.0 - 1.0
    if len(im.shape) == 3:
        im = im[None]  # expand for batch dim
    t2 = time_sync()
    dt[0] += t2 - t1
    # Inference
    # 預測
    pred = model(im, augment=augment, visualize=visualize)
    t3 = time_sync()
    dt[1] += t3 - t2
    # NMS
    pred = non_max_suppression(pred, conf_thres, iou_thres, classes, agnostic_nms, max_det=max_det)
    dt[2] += time_sync() - t3
    #用于存放結(jié)果
    detections=[]
    # Process predictions
    for i, det in enumerate(pred):  # per image 每張圖片
        seen += 1
        # im0 = im0s.copy()
        if len(det):
            # Rescale boxes from img_size to im0 size
            det[:, :4] = scale_coords(im.shape[2:], det[:, :4], im0.shape).round()
            # Write results
            # 寫入結(jié)果
            for *xyxy, conf, cls in reversed(det):
                xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4))).view(-1).tolist()
                xywh = [round(x) for x in xywh]
                xywh = [xywh[0] - xywh[2] // 2, xywh[1] - xywh[3] // 2, xywh[2],
                        xywh[3]]  # 檢測到目標位置，格式：（left，top，w，h）
                cls = names[int(cls)]
                conf = float(conf)
                detections.append({'class': cls, 'conf': conf, 'position': xywh})
    #輸出結(jié)果
    for i in detections:
        print(i)
    #推測的時間
    LOGGER.info(f'({t3 - t2:.3f}s)')
    return detections
path = 'C://Users//25096//Desktop//yoloV5//yolov5//yolov5-master//data//images//zidane.jpg'
img = cv2.imread(path)
#傳入一張圖片
detect(img)