Stuff Made Here on YouTube made this basketball hoop that tracks and tilts to make sure you never miss. Its fun to watch the process and trial and error! Thousands have told me that anyone can easily miss my first automatic basketball hoop by missing the hoop entirely. That is a really good point and something that I can't let stand. In this video I show you how I devoted several weeks of my life to realizing a basketball hoop that makes your shot go in even if you totally miss the hoop….

Engineer and YouTuber Shane Wighton has made a basketball hoop that uses a Microsoft Kinect and facial recognition in order to build a basketball hoop that means the shooter never misses. On the YouTube channel Stuff Made Here, Wighton explains that the backboard is tracking the information in the room, including the ball and its trajectory. With that information, the backboard can calculate where it needs to move in order to ensure the ball gets into the hoop. Since there are only 600 milliseconds (a thousandth of a second) between when the ball is thrown and when it hits the backboard, the calculations need to be made in an incredibly short amount of time. Therefore, Wighton said, when designing the board he had to prioritise fast movement. There are three motors, giving the machine three degrees of motion, and a universal joint connecting them to the board.

Now that my machine learning model can detect the different basketball artefacts, we can finally work on some basketball game logic. We can probably detect a goal by adding "region detection boxes" above and below the rim. The "top box" turns red when it detects a ball in its region above the rim. And the "bottom box" shows green when a ball is detected below the rim (where the basketball hoop is). Via the ML model we know with 92% - 94% accuracy where the basketball hoop is located so we can place the red/green boxes automatically on top of the video canvas.