Oh wow. Notably, Figures 7 and 8 (on pages 7 and 10, respectively) are better resolved than the stock photo.

Figure 8 shows a variety of "failure cases," but even these failure cases are surprisingly accurate estimations!

If I understand the captions correctly, the wireframe images overlaid on RGB images are from the algorithm run on the RGB images, not the wifi data, so having them match up is not surprising.

The standalone wireframe images sometimes manage to get the number of people and their approximate positions correct, but everything else about them seems (IMO) believable but not necessarily accurate, like most AI output.

This doesn't surprise me, because the input data is very coarse. Look at their graphs. There's just not enough raw data there to accurately reconstruct much other than position in the room.

I'm sure it's possible to do more accurate recording and reconstruction, I just don't think this demonstrates it.

Is that link broken for anyone else? Redirects back to the top level page for me

Try https://arxiv.org/pdf/2301.00250.pdf -- looks like they removed the original PDF from the linked article.