In a video on YouTube titled “Tesla Autopilot is better than you think! Here’s why,” Wes Bard gives a detailed analysis as to why this is so. In the video (featured above), Bard goes over the basics of what people think when they think of Autopilot. They think about what it can do, such as keeping you in a lane even while driving along a tight curve or adjusting the speed to match the car ahead of you. But this is just the tip of the iceberg, according to Bard.
He noted that the smallest details provide the key to seeing just how impressive Autopilot is. Bard used a video that Tesla released February of 2020 showing what the forward-facing camera sees to hone in on some of those details.
He started with the painted arrows on the road. In the video, they are labeled a certain way: LA, FA, and RA. These stand for Left Arrow, Front or Forward Arrow, and Right Arrow. If you look toward the horizon, you can see more of these labels in the distance. “If you’re not impressed yet, just remember that this is a computer recognizing arrows on a video that it has never seen before,” Bard said.
Bard pointed out that Autopilot first saw the stop sign “really far back” and zoomed in to show that Autopilot saw the stop sign outlined and labeled in red. As he drove closer toward the intersection, one important detail was the red line where the vehicle should come to a complete stop along with a label similar to those on the arrows — the letter S.
In another example with a truck in front of the vehicle, Bard remarked that Autopilot was still able to properly label the stop sign line on the road — whereas human eyes can’t see the physical line if it’s obscured by a vehicle.
A logical thinker would automatically know that place is where you should come to a complete stop since there is a stop sign. The fact that a computer knows this as well shows just how advanced at critical thinking while driving Autopilot has become.
Bard said that this was one of the most impressive things that he has seen Autopilot do. In the video, he pointed out the text on the left where you see a numerical value followed by text. If you notice in the screenshot above, most of the text is written in white except for the “Restricted” line.
Just above “Restricted” is the “Wet_Road” line. Keeping an eye on that as the video plays, you can see that as the car drives closer to a puddle just before the stop sign, the wet road line turns green or yellow. Also, those numbers change. “This is a computer recognizing water on a road from a video feed that it has never seen before,” Bard remarked.
Little Red X
If what Bard suspected the little red x means is accurate, then this is just more proof of how intelligent Autopilot has become. The little red x appears right where the road splits to go left or right. Bard suspects that the red x is a labeling of the right lane line as a lane line that Autopilot should not follow.
I also noticed something in the video — that little red x is just to the right of the right line while barely touching the white line — as if blocking an entrance. X usually means stop symbolically, and an intelligent person who knows this would see the red x and veer towards the left.
Other things that Bard covered in his video were traffic lights, labeling of other cars, indicating a lane line where cars are parked. “Even though there’s no actual line painted, the car recognizes that those parked cars are a border in a sense where the car cannot go and therefore puts a lane line.”
This is exciting to think about. I appreciate Bard’s analysis. When Tesla tweeted that video, I didn’t really pay attention to the details. To me, it looked like the inside of a computer game and something that was advanced in a way that I couldn’t understand.
Bard’s analysis helped my own mind to see more than just fast-moving, squiggly, colored lines dancing around the screen and I think this is what I appreciate the most from his video. He explained it in a way that the average person who may not have the mind of a tech person could understand it.
This makes me appreciate Autopilot even more. I mean, think about it. A computer can take information from its own environment and make split-second decisions based on the factors that are happening in real-time. I have to agree — Tesla’s Autopilot is better than we often think, and will continue to improve.
Have a tip for CleanTechnica, want to advertise, or want to suggest a guest for our CleanTech Talk podcast? Contact us here.