A brief sneak-peek at what we can expect to see for photography on the new iPhone 12 Pro.

简要介绍一下我们期望在新款iPhone 12 Pro上看到的摄影作品。

When the iPhone 11 Pro was announced everyone focused immediately on the new wide-angle lens which made complete sense as the new capability did give the new iPhone a big advantage over the previous generations. However, other capabilities such as portrait mode were given only minor updates and still to this day suffer from some seriously valid quality issues. If the new rumors are true, then the next generation iPhone 12 Pro will feature a new type of sensor that should drastically improve the quality of portrait mode, even to the point where professional photographers will take notice.

当iPhone 11 Pro发布时，每个人都立即专注于新的广角镜，因为新功能确实使新iPhone拥有比前几代产品更大的优势，这是完全有意义的。 但是，其他功能(例如人像模式)仅进行了较小的更新，直到今天仍然受到一些严重有效的质量问题的困扰。 如果新传闻属实，那么下一代iPhone 12 Pro将配备一种新型传感器，该传感器应可以大大改善人像模式的质量，甚至可以使专业摄影师注意到这一点。

With today’s iPhone 11 Pro you get significantly better quality images but we are still seeing the same edge detection problems that started with the iPhone 7 Plus.

使用今天的iPhone 11 Pro，您可以获得质量明显更好的图像，但是我们仍然看到与iPhone 7 Plus相同的边缘检测问题。

When Apple first announced portrait mode on the iPhone 7 Plus the technology was barely able to generate decent human portrait images and suffered from lots of poor edge detection issues. When the technology worked, it was amazing and it generated lots of awesome images but when it failed, it usually failed hard, and in ways that made the image unusable. Flash forward to today’s iPhone 11 Pro and while you get significantly better quality images for sure we are still seeing the same edge detection problems, especially when subjects are in front of a complex background. The reason for this is that the iPhone can’t actually generate a “real” depth map of the scene, it predicts a depth map from optical data that is run through a neural network. While this can still produce amazing results, it will still often fail at separating the subject from complex backgrounds that it has not been explicitly trained to understand. While training the neural network with more data will continue to make small improvements, Apple knows this approach won’t fix the problem for all possible situations. The real fix is to add a physical depth sensor and that’s exactly what they will be including on the iPhone 12 Pro.

当苹果公司首次在iPhone 7 Plus上宣布人像模式时，该技术几乎无法生成像样的人像图像，并且遭受了许多不良的边缘检测问题。 当这项技术奏效时，它是惊人的，并且产生了许多很棒的图像，但是当它失效时，它通常会很难失效，并以某种方式使图像无法使用。 快闪到今天的iPhone 11 Pro，尽管您可以得到质量更好的图像，但是可以肯定的是，我们仍然会遇到相同的边缘检测问题，尤其是当对象处于复杂背景下时。 原因是iPhone实际上无法生成场景的“真实”深度图，而是根据通过神经网络运行的光学数据预测深度图。 尽管这仍然可以产生惊人的结果，但仍然常常无法将主题与尚未经过明确训练以理解的复杂背景分开。 尽管用更多数据训练神经网络将继续做出微小的改进，但Apple知道这种方法不能在所有可能的情况下解决问题。 真正的解决方法是添加一个物理深度传感器，这正是它们将包括在iPhone 12 Pro中的功能。

From the first time I saw that Apple included a LiDAR sensors on the latest generation of iPad Pros I knew exactly how they were going to fix portrait mode on the iPhone. Using the LiDAR sensor to generate super-accurate depth maps will super-charge the iPhone’s ability to generate near-perfect depth-of-field control over the image capture process. No more edge detection errors and the ability to control depth-of-field no matter what the subject is will be a huge game changer. The ability to produce professional quality portraits, and to fully, and accurately, control depth-of-field will further close the gap between the iPhone and dedicated professional camera gear.

从我第一次看到苹果在最新一代的iPad Pro上安装了LiDAR传感器后，我就确切知道他们将如何在iPhone上修复人像模式。 使用LiDAR传感器生成超精确的深度图将使iPhone能够在图像捕获过程中生成近乎完美的景深控制能力。 无论对象是什么，都不会再有边缘检测错误和控制景深的能力，这将改变游戏规则。 产生专业品质肖像的能力以及完全，准确地控制景深的能力将进一步缩小iPhone与专用专业相机装置之间的差距。

In addition to improving the out-of-camera results for portrait mode the ability to generate accurate depth maps will also drastically improve the quality of predicted depth maps for app like Focus as they now have access to nearly unlimited training data from which to improve their neural networks. I predict that over time the ability to generate depth maps for virtually any image will become the normal way for photographers to manipulate depth-of-field and it will become increasingly rare to shoot using gear that does not let you change your virtual aperture (depth-of-field) during post-processing. The really big challenge to that transition is access to a reliable and accurate way to generate depth maps that don’t have visible errors. Once that’s solved, and it looks highly likely Apple will solve it using a LiDAR sensor on the iPhone 12, portrait mode will graduate from an interesting feature to something even professionals can rely on to create amazing images.

除了改善人像模式的相机外结果之外，生成准确深度图的能力还将大大提高像Focus之类的应用程序的预测深度图的质量，因为它们现在可以访问几乎无限的训练数据来改善其神经网络。 我预计随着时间的流逝，几乎可以为任何图像生成深度图的能力将成为摄影师操纵景深的常规方法，并且使用不允许您更改虚拟光圈(深度)的齿轮进行拍摄的情况将越来越少。场)。 过渡的最大挑战是获得可靠且准确的方式来生成没有可见错误的深度图。 一旦解决，苹果很有可能会使用iPhone 12上的LiDAR传感器来解决它，人像模式将从有趣的功能逐渐演变成甚至专业人士都可以用来创建惊人图像的功能。