这个机器人根据 Twitch 聊天命令抛硬币

做出决定是困难的，这就是为什么人类想出了各种愚蠢的方法来减轻一些精神负担。例如，优点/缺点列表应该尽可能增加决策过程的客观性

This internet-connected robot will flip a coin for you at the drop of a chat comment.

这个连接互联网的机器人会在您发表聊天评论时为您掷硬币。

Making decisions is hard, which is why humanity has come up with all sorts of goofy ways to alleviate some of the mental burden. Pros/cons lists, for example, are supposed to add as much objectivity as possible to the decision-making process — or at least to give you clarity about the choice you’re subconsciously rooting for. But the ultimate decision maker is obviously the good ol’ fashioned coin flip. You get 50:50 odds and the entertainment of watching something shiny tumble through the air.

做出决定是困难的，这就是为什么人类想出了各种愚蠢的方法来减轻一些精神负担。例如，优点/缺点列表应该为决策过程增加尽可能多的客观性，或者至少让你清楚地了解你潜意识支持的选择。但最终的决策者显然是传统的抛硬币。您将获得 50:50 的赔率，并享受观看闪亮物体在空中翻滚的乐趣。

To make coin flips accessible to his chat over the internet, Terence Grover built this robot.

为了让他可以通过互联网聊天进行抛硬币，特伦斯·格罗弗（Terence Grover）建造了这个机器人。

This robot lets anyone flip Grover’s very special coin via a simple command in a live stream chat. Commenting “!heads” flips the coin with heads as the prediction, and commenting “!tails” does the opposite. After receiving the command, the robot flips the coin. It then snaps a photo and analyzes that to determine which face is up. Finally, it displays the results and repositions the coin for the next flip. Grover also created a nifty graphic to display some basic statistics on the previous flips, including the heads:tails ratio and the percentage of correct predictions.

这个机器人可以让任何人通过实时流聊天中的简单命令来翻转格罗弗非常特别的硬币。评论“!heads”会以正面作为预测，而评论“!tails”则相反。机器人收到命令后，抛硬币。然后它会拍摄一张照片并对其进行分析以确定哪张脸朝上。最后，它显示结果并重新定位硬币以进行下一次抛掷。格罗弗还创建了一个漂亮的图形来显示之前翻转的一些基本统计数据，包括正面：反面比率和正确预测的百分比。

The heart of the machine is a small solenoid that provides the flipping actuation. If we have one criticism, it is that the solenoid seems weak and unable to give the coin a vigorous flip — though we’ll let the statisticians debate whether or not that affects the odds. A Raspberry Pi single-board computer controls that solenoid via a relay. It also controls a servo motor that opens and closes a 3D-printed iris mechanism, which is a clever way to recenter the coin over the solenoid after each flip.

该机器的核心是一个提供翻转驱动的小螺线管。如果我们有一个批评，那就是螺线管似乎很弱，无法使硬币有力地翻转——尽管我们会让统计学家争论这是否会影响赔率。 Raspberry Pi 单板计算机通过继电器控制该电磁阀。它还控制打开和关闭 3D 打印虹膜机构的伺服电机，这是每次翻转后将硬币重新置于螺线管上方的巧妙方法。

The Raspberry Pi runs a simple server that uses the GPIO pins to toggle the relay in response to a command received over the network, which is how it is able to integrate with the chat comments. It detects the coin face using a computer vision machine learning model trained on this specific coin in this machine, using images from a connected Raspberry Pi Camera. To speed up that process, it first crops most of the photo by detecting a circle of the approximate diameter of the coin. That makes the image very small and easy to process.

Raspberry Pi 运行一个简单的服务器，该服务器使用 GPIO 引脚来切换继电器以响应通过网络接收到的命令，这就是它能够与聊天评论集成的方式。它使用计算机视觉机器学习模型来检测硬币的正面，该模型是在该机器中针对该特定硬币进行训练的，并使用来自连接的 Raspberry Pi 相机的图像。为了加快这一过程，它首先通过检测大约硬币直径的圆圈来裁剪大部分照片。这使得图像非常小且易于处理。

A separate webcam watches the robot in action and streams the video to online viewers. That camera has a delay of a few seconds, but that matches up well with the computer vision processing time and so viewers see the detected results at about the same time as they see the corresponding flip in the video.

一个单独的网络摄像头监视机器人的动作，并将视频传输给在线观众。该相机有几秒钟的延迟，但这与计算机视觉处理时间很好地匹配，因此观看者在看到视频中相应翻转的同时几乎可以同时看到检测到的结果。

And to introduce a little extra drama, Grover gave the robot a special €2 coin celebrating some Monégasque prince. According to Grover, that is worth €1,000 and that raises the stakes of the coin flips.

为了引入一点额外的戏剧性，格罗弗给了机器人一枚特殊的 2 欧元硬币，以纪念摩纳加王子。据 Grover 称，这价值 1,000 欧元，这增加了抛硬币的赌注。