计算机视觉实战项目4（单目测距与测速+摔倒检测+目标检测+目标跟踪+姿态识别+车道线识别+车牌识别+无人机检测+A_路径规划+行人车辆计数+动物识别等）-2

        for path, img, im0s, vid_cap in dataset:
        img = torch.from_numpy(img).to(device)
        img = img.half() if half else img.float()  # uint8 to fp16/32
        img /= 255.0  # 0 - 255 to 0.0 - 1.0
        if img.ndimension() == 3:
            img = img.unsqueeze(0)

        # Warmup
        if device.type != 'cpu' and (old_img_b != img.shape[0] or old_img_h != img.shape[2] or old_img_w != img.shape[3]):
            old_img_b = img.shape[0]
            old_img_h = img.shape[2]
            old_img_w = img.shape[3]
            for i in range(3):
                model(img, augment=opt.augment)[0]

        # Inference
        t1 = time_synchronized()
        with torch.no_grad():   # Calculating gradients would cause a GPU memory leak
            pred = model(img, augment=opt.augment)[0]
        t2 = time_synchronized()
         distance=object_point_world_position(u, v, h, w, out_mat, in_mat):

###############创建A-Star类############
class AStar:
    
    # 描述AStar算法中的节点数据
    class Node:  
        #初始化
        def __init__(self, point, startPoint,endPoint, g=0,w=1,p=1):
            self.point = point  # 自己的坐标
            self.father = None  # 父节点
            self.g = g       # g值，g值在用到的时候会重新算
            
            # 计算h值，采用曼哈顿距离
            #self.h = (abs(endPoint.x - point.x) + abs(endPoint.y - point.y)) * 10  
            
            #采用欧几里得距离
            #self.h = math.pow((math.pow((endPoint.x - point.x),2) + math.pow((endPoint.y - point.y),2)),0.5)*10
            
            #采用对角距离
            pp=(1-p)+0.2*math.exp((math.pow((math.pow((endPoint.x - point.x),2) + math.pow((endPoint.y - point.y),2)),0.5))/(math.pow((math.pow((endPoint.x - startPoint.x),2) + math.pow((endPoint.y - startPoint.y),2)),0.5)))
            Diagonal_step = min((endPoint.x - point.x),(endPoint.y - point.y))
            straight_step = (abs(endPoint.x - point.x) + abs(endPoint.y - point.y)) - 2*Diagonal_step
            self.h  =(straight_step + math.pow(2,0.5)*Diagonal_step)*10*pp
            #print(pp)


            
    #初始化A-start
    def __init__(self, map2d, startPoint, endPoint, passTag=1.0):#map2d地图信息，startPoint起点, endPoint终点, passTag=1.0为不可行驶区域

        # 开启表
        self.openList = []
        # 关闭表
        self.closeList = []
        # 寻路地图
        self.map2d = map2d
        # 起点终点
        if isinstance(startPoint, Point) and isinstance(endPoint, Point):
            self.startPoint = startPoint
            self.endPoint = endPoint
        else:
            self.startPoint = Point(*startPoint)
            self.endPoint = Point(*endPoint)
 
        # 不可行走标记
        self.passTag = passTag
 
    def getMinNode(self):
        """
        获得openlist中F值最小的节点
        :return: Node
        """
        currentNode = self.openList[0]
        for node in self.openList:
            if node.g + node.h < currentNode.g + currentNode.h:
                currentNode = node
        return currentNode#返回最小代价的点
 

#全部代码可加qq1309399183
def train():
    #parses command line args
    args = parse_args()

    #parses args from file
    if args.config_file is not None:
        cfg_from_file(args.config_file)

    if (args.FIX_MODEL_CHECKPOINT):
      args.FIX_MODEL_CHECKPOINT = args.FIX_MODEL_CHECKPOINT.replace(" ", "")
      args.FIX_MODEL_CHECKPOINT = args.FIX_MODEL_CHECKPOINT.replace("=", "")
      cfg.RESUME_CHECKPOINT = args.FIX_MODEL_CHECKPOINT
      cfg.CHECK_PREVIOUS = False
      if (os.path.exists(cfg.RESUME_CHECKPOINT) == False):
          print('Exiting the process as asked model for resuming is not found')
          exit()

    if (args.RESUME_CHECKPOINT):
      cfg.RESUME_CHECKPOINT = args.RESUME_CHECKPOINT

    if (args.LOG_DIR):
      cfg.EXP_DIR = args.LOG_DIR

    cfg.LOG_DIR = cfg.EXP_DIR

    if (args.PHASE):
      cfg.PHASE = []
      cfg.PHASE.append(args.PHASE)

    if (args.EVAL_METHOD):
      cfg.DATASET.EVAL_METHOD = args.EVAL_METHOD

    #for backward compatibility
    if cfg.DATASET.DATASET == 'psd':
      cfg.DATASET.DATASET = 'tiod'

    if cfg.DATASET.BGR_OR_RGB == True:
        #cfg.DATASET.PIXEL_MEANS = (123.68, 116.78, 103.94)
        #cfg.DATASET.PIXEL_MEANS = (123, 117, 104)
        cfg.DATASET.PIXEL_MEANS = (128.0, 128.0, 128.0) # simpler mean subtraction to keep data in int8 after mean subtraction

    print("cfg: ", cfg)

    for phase in cfg.PHASE:
      cfg_dir = cfg.LOG_DIR + '/' + phase + '_cfg/'
      os.makedirs(os.path.dirname(cfg_dir), exist_ok=True)
      shutil.copy(args.config_file, cfg_dir)

    # to making every run consistent # TII
    np.random.seed(100)
    torch.manual_seed(100)
    torch.cuda.manual_seed(100)
    random.seed(100)
    torch.cuda.manual_seed_all(999)
    torch.backends.cudnn.enabled = False

    train_model()

if __name__ == '__main__':
    train()