I don’t know if this can be seen as a night photo.
But with my smartphone I can shoot in night mode and see things I cannot see with the naked eye.
Of course, I cannot move.
I don’t think the focus is the problem. In those pictures the far away parts of the picture are kind of sharp. Focusing to infinity would achieve the same, if that was the problem. I think it is more like, the far away parts do not move as quickly and therefore are less affected by the motion blur.
About general capture at night: For the following I uses OpenCamera, which allows manual settings, specifcally it allows to set the shutter speed and the ISO:
It worked, and it produced usable images. A bit grainy and fuzzy due to high ISO and strong postprocessing, but usable for mapping, sign recognition etc.
However I needed to adjust that manually all the time as it got darker. This could be easily provided automatically, but android somehow doesn’t want to: Shutter Priority?
I also use OpenCamera. Its not clear to me how to get Mapillary to use it. Is it a matter of defaults?
The general guidance is not to shoot in low light/night if at all possible. If you really want to because this is an area that has no coverage and you know what you’re doing, you can use OpenCamera and upload the images with the Desktop Uploader.
1 Like
They removed the focus-to-infinity feature a while back, which has caused me to not use Mapillary as much now. That was the best focus mode to use.
2 Likes
Although not working on all smartphone, the focus-to-infinity option was a nice option to get smoother acquisition.
BTW, my pictures made with DSLR were recorded with preset manual focus (but camera was just placed on the dashboard, not firmly fixed).
Hi folks - we didn’t intentionally remove focus-to-infinity, in fact the camera should be set to focus-to-infinity when you are moving quickly (not on foot). I see this isn’t working as expecting though - we are going to investigate a fix.
2 Likes
The three vertical dots, then you see the photo modes.
Try it in an almost dark room, if you have patience, you will be amazed.
Normally one can see better in real life than on a photo. Here it is the contrary.
I do plan on using my tripod later. So far, I have just taken a few test sequences.
Is this guidance solely based on image quality?
I ask because, as time goes by, camera technology is improving. In 2015 (when the thread was made), I would not even consider uploading a night-time action camera footage. But now in 2024, I look at some of the smoother nightime sections and go “huh, this is almost passable”. Like, people’s shaky phone sequences are worse than my nighttime action shots. So it’s now becoming relevant and relative. For certain, I can record later into the day with less noise and blur while sun glare / sunset angle no longer “destroy” the footage. I imagine if I had a proper gimbal at night, I could reduce blur by a lot.
For context, here’s a (25%-sized) video still from a more-or-less stationary GoPro 12 from a really dark night:
For something like OSM perspective, this helps with mapping street lights and lit streets and such. But one big reason why I don’t upload non-day photos though is that Mapillary has no way to filter them afterwards. Even if I wanted to map lit streets, for example, I would have no way to find relevant sequences later.
3 Likes
Adding filtering by time is an interesting idea - so that you could filter for night photos if that’s really what you wanted.
In general night photos are less desirable because image quality impacts everything else (your ability to see house numbers, read signs, detect traffic signs accurately, etc). Low light also generally leads to longer exposures which can lead to photo blurriness. So there are cases where you could do evening photography for a specific purpose, but as a general rule of thumb it is not recommended.
1 Like
@boris I see some contributors dump more and more low quality night time imagery. Imho, this kind of imagery should be filtered and rejected based on “Terms of Use”. Low quality imagery messes up reconstruction, obstructs other useful day time sequences on the map, and generally adds little to no value but just wastes scarce computing resources, energy, and money. When mapping some places, I am really fed up with the situation as it currently is. In the age of generative and generational AI it should not be that hard to compute a score and just reject low quality imagery below a certain threshold. Mapillary has started with a clear mission statement to “capture street level imagery for mapping the world” and everybody should adhere to that. Mapillary is not and should not be a dumpster. Trying to be inclusive by accepting everything below every standard just defeats every purpose. Nobody should want to accept such reasoning and behavior. One reason why the competition, e.g. Google Maps, is so successful is because they have standards they adhere to. I genuinely wish that Mapillary takes this seriously and does something about it.
2 Likes
Thank you for sharing your thoughts as always @GITNE !
Agreed that Mapillary is not and should not be a dumpster 
I think there are smart things we can do to showcase the best imagery more prominently (might be a combination of things like lighting condition, camera view angle, resolution, obstructions, recency, sharpness, etc). It’s something we’ve been thinking a bit about, and will plan to make make some steps in that direction this year.
Allow me to deviate a bit from the street, as a reward to me.
1 Like