Flying Backwards Through Time

I've tried to make this video as self-explanatory as possible, so here it is without further ado :

This will be a short post just to give a bit more detail and some static images.

I was feeling really inspired and motivated by the recent solarigraphy meeting. I haven't made any dedicated data visualisation projects in a while, but one that's been gestating in the back of my mind is to try and show the dark matter in galaxy simulations... I think showing this as trails (a la the Sun in solarigraphic images) might look interesting.

But first, there's much lower-hanging fruit to pluck, though in a tangentially similar vein. Back in 2018, I decided it might be fun to try a time lapse from the office. Specifically, here, looking down the Jihozápadní V avenue.

Red X marks the spot.

So each time I went to the office, for the next ~16 months, I took a photo on my mobile phone, which was then a Huawei P10 lite. I tried to keep things as stable as I could, trying to get myself in the same position each time and lining things up in a similar way, but this was all done very quickly and by hand. I decided to do this particular view basically on a whim : it's somewhere I was present regularly and reliably at about the same time, it's quite a nice view, and looking in the other direction (towards the institute itself) you don't see so much. From the other direction, where you enter the avenue to go towards the institute, at certain times of year one gets a very nice "Stonehenge" effect :

Looking in the other direction avoids the sun being in the frame and I'm not sure if this was the right decision or not. I do wish I'd gone for landscape format, but having to avoid the tops of the trees would probably have annoyed me quite intensely.

Anyway, I didn't always manage to take a photograph every day. Sometimes I just forgot, occasionally going back later when I remembered (or if there were inconveniently-placed people that I hoped would bugger off if I left it for a bit). Sometimes I was away for conferences or holidays* (maybe even working from home from time to time !), and of course I wasn't there on weekends at all. Also the time of day varied considerably, from 7:45 am (God knows what that was all about) to 5:36 pm (probably I forgot to take the picture in the morning). The median time was 9:47 am, with a standard deviation of 1 hour. 

* Worst of all for this project, I tend to take holidays in the spring an autumn, so the most crucial moments of seasonal change are missing !

This is about as good as can be expected from a real-world side-project, where it's just not possible to keep things more controlled. So not only does the camera position vary quite a bit, but so does the lighting - and of course, this varies dramatically depending on the weather.

The result is the first sequence in the video - not much more than an unpleasant mess. We need not dwell on it any longer.

I was expecting as much. I knew I'd probably have to manually align the images to get a nice result, but somehow I just never got around to doing it... until now. So I imported all the images in Blender as transparent planes and manually aligned them by adjusting their scaling, rotation and location. If we show all the images together, here's the unaligned result :

Which explains why the animated sequence is such a mess. Here's the version after doing all the corrections :

Little bit of contrast adjustment on this one.

Much better. Probably there are better ways to do this; certainly there are a few cases where I didn't align things as well as I could, probably it could even be automated. But I enjoyed the process, so never mind. I'm also pleased with the way the colour turned out, and the blur gives it an air of authenticity, like it's from an archaic film camera instead of a modern electronic gizmo. Still, while the resulting animation sequence is far more bearable than the raw version, it's not quite as good as I was hoping for.

But together with all the long-exposure stuff I'd seen in the solarigraphy meeting, this gave me ideas of how to process the images differently. Why stop with a simple, traditional sequence of images ? These combined images are in effect pseudo-long-exposure photographs, compressing 16 months into single stills. They're not the same as proper solarigraphs of course, but the overall effect is similar in that you see multiple times all at once.

So I tried combining the images in different ways for the animation. The simplest approach, which gives a result I quite like, is to give all the images the same level of transparency, such that only when they're all combined do they become completely opaque. That's what's shown in still images above. In the animation, it gives very a much smoother animation than the original, and you still get a satisfying sense of change as it progresses. 

But why insist we see all the images ? If we restrict it so that only 30 images (one month) are ever shown on screen, and their transparency is such that their sum is fully opaque, then we get the best of both worlds : a nice smooth transition but with a more pronounced sense of change.

Choosing four images to give an obligatory "four season" compilation. We had a particularly hot and dry summer that year, with all the grass dead and even the trees looking unhealthy. Fortunately we also had a cold, snowy winter.

Finally, using the same technique I do for astronomy data visualisations, we can also show the data as a pseudo-3D volume. In effect this has been done already : the images are all separated along the into-the-screen axis, it's just you can't see this because the rendering disables perspective. If we enable it, we get this funky volumetric effect. I really like this in the animation, you get a sense of movement that's only partially true... the camera moves back from the images, but that's not the same as really moving through the street. It's actually moving backwards through time, a weird quasi-movement away from the images but not through the street itself, all the while showing images from longer and longer ago.

Or equivalently, in this image you're looking into the future by looking at the centre, and into the past by looking at the edges. Time is treated as a spatial dimension.

This concludes my brief foray into experimental photography. You may go about your business, and I'll get back to the world of astronomy.