Totally clickbait the idea of testing out different lenses is great but in order to see behind the wall he just puts the camera higher around the wall and then makes the picture.
I was personally expecting a breakthrough or a mechanism that is smaller and doesn't go above the object in order to make a picture from the simulated lenses.
The article doesn’t explain why this is interesting. Here’s a different way of putting it:
A classic pinhole camera is a box with a hole in it. The hole is small, so not much light gets in. We replace it with a lens that acts in effectively the same way, but lets in more light. The pinhole is inside the camera, and there is an obstacle in front of the subject, so the camera can’t see it. The fun thing here, is that this camera has a pinhole _outside_ the camera, so the pinhole can be positioned on the other side of the obstacle, and therefore take a picture of the subject.
You can visualize it easily if you think about your own vision.
If you put an hand in from of your face ( without covering your eyes ) you will be able to see behind it even if both eyes see only a part of what is behind your hand.
Now regarding the video imagine that each pixel is an eye, and they are spreaded evenly along a circle.
There are a lot of differences between this example and what he actually did, but it should be very easy to visualize ( main difference I can think of is how much amplification he needed to do so each eye is almost blind )
Same reaction here but with a little more thought I give it more credit. The resulting image looks like you're looking at the subject head-on from a point behind the wall. It doesn't look like you're peeking over the wall at some angle.
The actual video doesn't have this headline, the clickbait comes from this article that doesn't really add anything over the video. It's clearly referencing the part of the video where says "sees around walls", but actually that is more accurate anyway compared to "sees behind walls" because it does see all around the wall. The ring lens characterization in the video is even better at not setting a weird expectation.
The video itself is wonderful regardless. I thought the ignored and reversed perspective shots were amazing.
Notice his device is always facing forward and it doesn’t turn to side to get pixel data so you would expect to get an image from what is exactly in front of the sensor. But it is not in case where he sets it when making picture “behind”.
The emulating different lenses part is that with this setup you can parametrically collect specific rays bouncing off the object (presumably the angle of the mirror with respect to radius), and the resulting image will be the same as you’d get from a lens with those parameters.
