A Youtuber has developed and built his own traffic sign recognition - it works better than Tesla's.
Traffic sign recognition now works tolerably, but without working completely error-free. Tesla drivers in particular have often complained about the insufficient capabilities of the sign recognition. The Hamburg YouTuber and Tesla driver Leo Tiedt also had bad experiences with the technology – and thought that there could be a better solution. Thought, said, done - the 21-year-old hobbyist is really launching his own image recognition. The hobby programmer does not have any specialized training.
Lone fighter against factory programming team
For its traffic sign recognition, Tiedt trains machine learning algorithms, i.e. software with so-called artificial intelligence, to recognize the German forest of signs. The technology should then simply display the recognized signs on a small additional monitor. The Youtuber wants to prove to Tesla that he, as an individual, can develop traffic sign recognition that is superior to Tesla's.
To Tiedt's amazement, there is a lot of reference material on the internet when it comes to machine learning. He also found a database of 50,000 images of German traffic signs that is specifically designed to add sign recognition to machine-learning software. After several days of data merging, his file with the martial name "der_Tesla_Zerstorer.pt" is ready.
Tiedt's system clearly wins
Tiedt's sign recognition technology consists of three components: A Raspberry Pi single-board computer, which is connected to an external camera from Logitech and a round LCD display - the display is not much larger than a Two euro piece. For his test drives, the YouTuber attaches the camera to the windshield with a suction cup near the interior mirror. The caseless Raspberry Pi sits on the center console. Because the small computer heats up when processing the large amounts of image data, Tiedt has also installed a classic fan that blows cooling air over the microcomputer.
The test drive takes place in bright sunshine and otherwise with the best visibility. Tiedt's self-made technology recognizes the first two signs, a construction site sign and a priority sign in one place. The following 60 km/h sign on the country road is also no problem for the detection system. Then it's off to the autobahn, where Tesla technology has particularly big problems. Tiedt's development didn't get any better at first either: the first 120 km/h speed limit sign went unnoticed. However, his system recognizes the following signs in rows and also considerably more often than the Tesla technology.
Tesla has detention
Leo Tiedt closes his video with a call to the Tesla developers to please take a few days to improve the sign recognition. If he can do that, then the Tesla developers should definitely be able to do that, concludes the Youtuber.
The Hamburg YouTuber and Tesla driver Leo Tiedt is dissatisfied with the Tesla traffic sign recognition, which is almost useless on German autobahns. So the hobby programmer built his own sign recognition with simple means within a few days and tried it out. The result is shameful for Tesla: Tiedt's system does not work perfectly, but far better than that of the American electric car manufacturer.
Tiedt deserves respect for his good work - and Tesla should finally improve the sign recognition issue for the benefit of its customers.