Working with the NSOperationQueue Class

Multi-tasking prevents apps from freezing. In most programming languages, achieving this is a bit tricky, but the NSOperationQueue class in iOS makes it easy!

This tutorial will demonstrate how to use the NSOperationQueue class. An NSOperationQueue object is a queue that handles objects of the NSOperationclass type. An NSOperation object, simply phrased, represents a single task, including both the data and the code related to the task. The NSOperationQueue handles and manages the execution of all the NSOperation objects (the tasks) that have been added to it. The execution takes place with the main thread of the application. When an NSOperation object is added to the queue it is executed immediately and it does not leave the queue until it is finished. A task can be cancelled, but it is not removed from the queue until it is finished. The NSOperation class is an abstract one so it cannot be used directly in the program. Instead, there are two provided subclasses, the NSInvocationOperation class and theNSBlockOperation class. I'll use the first one in this tutorial.

Here's the goal for this tutorial: for each extra thread we want our application to create an NSInvocationOperation (NSOperation) object. We'll add each object into the NSOperationQueue and then we're finished. The queue takes charge of everything and the app works without freezing. To demonstrate clearly the use of the classes I mentioned above, we will create a (simple) sample project in which, apart from the main thread of the app, we will have two more threads running along with it. On the first thread, a loop will run from 1 to 10,000,000 and every 100 steps a label will be updated with the loop's counter value. On the second thread, a label's background will fill with a custom color. That process will take place inside a loop and it will be executed more than once. So we will have something like a color rotator. At the same time, the RGB values of the custom background color along with the loop counter's value will be displayed next to the label. Finally, we will use three buttons to change the view's background color on the main thread. These tasks could not be executed simultaneously without multi-tasking. Here is a look at the end result:

demo_app

Let's begin by creating the project. Open the Xcode and create a new Single View Application.

project_template

Click on Next and set a name for the project. I named it ThreadingTestApp. You can use the same or any other name you like.

project_name

Next. complete the project creation.

project_creation

Click on the ViewController.xib file to reveal the Interface Builder. Add the following controls to create an interface like the next image:

ib_sample
  1. UINavigationBar
    • Frame (x, y, W, H): 0, 0, 320, 44
    • Tintcolor: Black color
    • Title: "Simple Multi-Threading Demo"
  2. UILabel
    • Frame (x, y, W, H): 20, 59, 280, 21
    • Text: "Counter at Thread #1"
  3. UILabel
    • Frame (x, y, W, H): 20, 88, 280, 50
    • Background color: Light gray color
    • Text color: Dark gray color
    • Text: -
  4. UILabel
    • Frame (x, y, W, H): 20, 154, 280, 21
    • Text: "Random Color Rotator at Thread #2"
  5. UILabel
    • Frame (x, y, W, H): 20, 183, 100, 80
    • Background color: Light gray color
    • Text: -
  6. UILabel
    • Frame (x, y, W, H): 128, 183, 150, 80
    • Text: -
  7. UILabel
    • Frame (x, y, W, H): 20, 374, 280, 21
    • Text: "Background Color at Main Thread"
  8. UIButton
    • Frame (x, y, W, H): 20, 403, 73, 37
    • Title: "Color #1"
  9. UIButton
    • Frame (x, y, W, H): 124, 403, 73, 37
    • Title: "Color #2"
  10. UIButton
    • Frame (x, y, W, H): 228, 403, 73, 37
    • Title: "Color #3"

For the last UILabel and the three UIButtons, set the Autosizing value to Left - Bottom to make the interface look nice on the iPhone 4/4S and iPhone 5, just like the next image:

autosizing

In this next step we will create the IBOutlet properties and IBAction methods that are necessary to make our sample app work. To create new properties and methods, and connect them to your controls while being the Interface Builder, click on the middle button of the Editor button at the Xcode toolbar to reveal the Assistant Editor:

assistant_editor

Not every control needs an outlet property. We will add only one for the UILabels 3, 5, and 6 (according to the order they were listed in step 2), named label1, label2, and label3.

To insert a new outlet property, Control+Click (Right click) on a label > Click on the New Referencing Outlet > Drag and Drop into the Assistant Editor. After that, specify a name for the new property, just like in the following images:

outlet1Inserting a new IBOutlet property

outlet2
Setting the IBOutlet property name

Repeat the process above three times to connect the three UILabels to properties. Inside your ViewController.h file you have these properties declared:

Now add the IBAction methods for the three UIButtons. Each one button will change the background color of the view. To insert a new IBAction method, Control+Click (Right click) on a UIButton > Click on the Touch Up Inside > Drag and Drop into the Assistant Editor. After that specify a name for the new method. Take a look at the following images and the next snippet for the method names:

ibaction1
Inserting a new IBAction method
ibaction2
Setting the IBAction method name

Again, repeat the process above three times to connect every UIButton to an action method. The ViewController.h file should now contain these:

The IBOutlet properties and IBAction methods are now ready. We can now begin coding.

One of the most important tasks we must do is to declare a NSOperationQueueobject (our operation queue), which will be used to execute our tasks in secondary threads. Open the ViewController.h file and add the following content right after the @interface header (don't forget the curly brackets):

Also, each task needs to have at least one method which contains the code that will run simultaneously with the main thread. According to the introductory description, the first task the method will be named counterTask and the second one will be named colorRotatorTask:

That's all we need. Our ViewController.h file should look like this:

Let's move on to implementation.

We're almost finished. We have setup our interface, made all the necessary connections, declared any needed IBAction and other methods, and established our base. Now it is time to build upon them.

Open the ViewController.m file and go to the viewDidLoad method. The most important part of this tutorial is going to take place here. We will create a newNSOperationQueue instance and two NSOperation (NSInvocationOperation)objects. These objects will encapsulate the code of the two methods we previously declared and then they will be executed on their own by the NSOperationQueue. Here is the code:

This whole process is really simple. After creating the NSOperationQueue instance, we create an NSInvocationOperation object (operation). We set its selector method (the code we want executed on a separate thread), and then we add it to the queue. Once it enters the queue it immediately begins to run. After that the operation object can be released, since the queue is responsible for handling it from now on. In this case we create another object and we'll use it the same way for the second task (colorRotatorTask).

Our next task is to implement the two selector methods. Let's begin by writing thecounterTask method. It will contain a for loop that will run for a large number of iterations and every 100 steps the label1's text will be updated with the current iteration's counter value (i). The code is simple, so here is everything:

Please note that it is recommended as the best practice (even by Apple) to perform any visual updates on the interface using the main thread and not by doing it directly from a secondary thread. Therefore, the use of the performSelectorOnMainThreadmethod is necessary in cases such as this one.

Now let's implement the colorRotatorTask method:

You can see that we used the performSelectorOnMainThread method here as well. The next step is the [NSThread sleepForTimeInterval:0.4]; command, which is used to cause some brief delay (0.4 seconds) in each loop execution. Even though it is not necessary to use this method, it is preferable to use it here to slow down the background color's changing speed of the label2 UILabel (our color rotator). Additionally in each loop we create random values for the red, green, and blue. We then set these values to produce a custom color and set it as a background color in the label2 UILabel.

At this point the two tasks that are going to be executed at the same time with the main thread are ready. Let's implement the three (really easy) IBAction methods and then we are ready to go. As I have already mentioned, the three UIButtons will change the view's background color, with the ultimate goal to demonstrate how the main thread can run alongside the other two tasks. Here they are:

That's it! Now you can run the application and see how three different tasks can take place at the same time. Remember that when the execution of NSOperation objects is over, it will automatically leave the queue.

Many of you may have already discovered that the actual code to run a multi-tasking app only requires a few lines of code. It seems that the greatest workload is implementing the required methods that work with each task. Nevertheless, this method is an easy way to develop multi-threading apps in iOS.

转载于:https://www.cnblogs.com/lisa090818/p/4253053.html

  • 0
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
1 目标检测的定义 目标检测(Object Detection)的任务是找出图像中所有感兴趣的目标(物体),确定它们的类别和位置,是计算机视觉领域的核心问题之一。由于各类物体有不同的外观、形状和姿态,加上成像时光照、遮挡等因素的干扰,目标检测一直是计算机视觉领域最具有挑战性的问题。 目标检测任务可分为两个关键的子任务,目标定位和目标分类。首先检测图像中目标的位置(目标定位),然后给出每个目标的具体类别(目标分类)。输出结果是一个边界框(称为Bounding-box,一般形式为(x1,y1,x2,y2),表示框的左上角坐标和右下角坐标),一个置信度分数(Confidence Score),表示边界框中是否包含检测对象的概率和各个类别的概率(首先得到类别概率,经过Softmax可得到类别标签)。 1.1 Two stage方法 目前主流的基于深度学习的目标检测算法主要分为两类:Two stage和One stage。Two stage方法将目标检测过程分为两个阶段。第一个阶段是 Region Proposal 生成阶段,主要用于生成潜在的目标候选框(Bounding-box proposals)。这个阶段通常使用卷积神经网络(CNN)从输入图像中提取特征,然后通过一些技巧(如选择性搜索)来生成候选框。第二个阶段是分类和位置精修阶段,将第一个阶段生成的候选框输入到另一个 CNN 中进行分类,并根据分类结果对候选框的位置进行微调。Two stage 方法的优点是准确度较高,缺点是速度相对较慢。 常见Tow stage目标检测算法有:R-CNN系列、SPPNet等。 1.2 One stage方法 One stage方法直接利用模型提取特征值,并利用这些特征值进行目标的分类和定位,不需要生成Region Proposal。这种方法的优点是速度快,因为省略了Region Proposal生成的过程。One stage方法的缺点是准确度相对较低,因为它没有对潜在的目标进行预先筛选。 常见的One stage目标检测算法有:YOLO系列、SSD系列和RetinaNet等。 2 常见名词解释 2.1 NMS(Non-Maximum Suppression) 目标检测模型一般会给出目标的多个预测边界框,对成百上千的预测边界框都进行调整肯定是不可行的,需要对这些结果先进行一个大体的挑选。NMS称为非极大值抑制,作用是从众多预测边界框中挑选出最具代表性的结果,这样可以加快算法效率,其主要流程如下: 设定一个置信度分数阈值,将置信度分数小于阈值的直接过滤掉 将剩下框的置信度分数从大到小排序,选中值最大的框 遍历其余的框,如果和当前框的重叠面积(IOU)大于设定的阈值(一般为0.7),就将框删除(超过设定阈值,认为两个框的里面的物体属于同一个类别) 从未处理的框中继续选一个置信度分数最大的,重复上述过程,直至所有框处理完毕 2.2 IoU(Intersection over Union) 定义了两个边界框的重叠度,当预测边界框和真实边界框差异很小时,或重叠度很大时,表示模型产生的预测边界框很准确。边界框A、B的IOU计算公式为: 2.3 mAP(mean Average Precision) mAP即均值平均精度,是评估目标检测模型效果的最重要指标,这个值介于0到1之间,且越大越好。mAP是AP(Average Precision)的平均值,那么首先需要了解AP的概念。想要了解AP的概念,还要首先了解目标检测中Precision和Recall的概念。 首先我们设置置信度阈值(Confidence Threshold)和IoU阈值(一般设置为0.5,也会衡量0.75以及0.9的mAP值): 当一个预测边界框被认为是True Positive(TP)时,需要同时满足下面三个条件: Confidence Score > Confidence Threshold 预测类别匹配真实值(Ground truth)的类别 预测边界框的IoU大于设定的IoU阈值 不满足条件2或条件3,则认为是False Positive(FP)。当对应同一个真值有多个预测结果时,只有最高置信度分数的预测结果被认为是True Positive,其余被认为是False Positive。 Precision和Recall的概念如下图所示: Precision表示TP与预测边界框数量的比值 Recall表示TP与真实边界框数量的比值 改变不同的置信度阈值,可以获得多组Precision和Recall,Recall放X轴,Precision放Y轴,可以画出一个Precision-Recall曲线,简称P-R
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值