Low‑Light “Supernight‑Style” Video Modes for Street‑Level Mapping – Surprisingly Useful Even in Daylight?

Contributing and equipment
1

Hi everyone,

I’ve been experimenting with different video modes for street‑level mapping and recently made an interesting discovery that I’d like to discuss with the community. Many 360° cameras offer a “Supernight” or low‑light optimized video mode. These modes are officially intended for night scenes, but I started testing whether they might actually be beneficial for mapping in general — including in bright daylight.

Here are my observations so far.

Daylight performance
I recorded several test drives in full sunshine using a low‑light optimized mode. Surprisingly, I did not see any degradation in image quality. Instead, I noticed:

• Very smooth transitions between bright and shaded areas
• No exposure pumping when exiting tunnels
• More stable overall exposure compared to standard video modes

So far, I haven’t found any downside when using this mode during the day.

High‑speed performance
I also tested this mode at high speeds (>100 km/h). Even at these speeds:

• I did not observe noticeable motion blur
• Road markings and traffic signs remained clear
• The footage looked stable and suitable for frame extraction

This was unexpected, because low‑light modes often use longer exposure times.

Low‑light and dusk performance
During a drive into advanced dusk, the mode continued to produce bright, usable footage even when it was already quite dark for the human eye. I observed:

• Clear traffic signs
• A slightly softened road surface (likely due to noise reduction)
• No sudden ISO jumps or exposure pumping
• Overall stable and consistent frames

This suggests that such modes allow mapping much deeper into the night than standard video settings.

General hypothesis
Based on these tests, I’m starting to think that low‑light optimized “Supernight‑style” modes might be the most stable and reliable option for street‑level mapping — not only at night, but also during the day.

Potential advantages:

• Very stable exposure
• Smooth transitions between lighting conditions
• Clear signage even in low light
• Minimal motion blur at high speeds
• Usable footage far into the night
• Consistent frames for extraction

For mapping platforms, these characteristics may be more important than maximum per‑frame sharpness.

Questions for the community
I’d love to hear your thoughts:

• Has anyone else experimented with low‑light or “Supernight” modes for mapping?
• Do you see similar behavior across different 360° cameras?
• Are there technical downsides I might be overlooking (compression, dynamic range, stitching, etc.)?
• How do these modes compare to standard video settings in your workflows?

Looking forward to your insights.

2

Honestly, I am not so super fond of night time modes for mapping. If you just want to take nice pictures on your holidays or you have some other special use case for it than it may be okay.

When mapping, one generally wants to capture reality, the ground truth, not some digital magic pixel dust. This is also why large scale professional street‑level and aerial/satellite mappers like Google Maps, Apple, tomtom, Maxar, etc. continue to invest heavily into dedicated, often purpose built, capture hardware. They could use lots and lots of cheap consumer grade Insta 360s, MAXes, etc. but they do not. So, why do not they? They do not do it because a) they need a large scale workflow and b) because of optical performance.

Night time modes are basically the top league of digital trickery. Everything and anything goes. The base premise of night time modes is to quickly capture multiple images at high ISO speeds. Then these images are combined into one final image after a cascade of digital filters. Usually, three (often even more) basic digital filter effects are used:

  • motion vector detection and pattern matching to reconstruct multiple images into one final image and to compensate for motion blur
  • digital de‑noising (due to high ISO speeds which naturally introduce lots of noise)
  • sharpening, gamma balancing, and color widening through interpolation of adjacent pixel values

All of the above has only become possible due to the miniaturization of general purpose computing GPUs. Otherwise, nobody would bother with night time modes because you would have to build costly dedicated chip circuitry. Does it work? Sure, sometimes even surprisingly well for the amount of resources available in such consumer grade cameras or smartphones. Do I want to use it for mapping? No, thanks.

I do. You just have to zoom in. :face_with_monocle: It comes through even with Mapillary garbling the original image quality. Noise and the cheapo de‑nosier’s desperate attempts to reduce that noise are also clearly visible. However, since you seem to be keen on capturing tunnels from a car in motion then a night time mode may indeed be useful for this particular scenario. Apart from that, I would rather recommend to stop capturing after sunset because usually nobody is really well served with murky imagery.

Since Mapillary has not really enforced any quality standards or levels yet, you should be able to get by with this. :person_shrugging: Like you say, maybe AI extractability is the lowest bar image quality has to pass for Mapillary?

3

The evening shot I chose comes from a situation where I had to stop recording shortly afterward due to increasing darkness. This example is already well beyond the point where I usually stop taking 360-degree photos. I’m interested in the situation in narrow street canyons where light and shadow alternate rapidly. Since you often can’t choose tunnel situations, I’m wondering if Supernight is the better recording mode for me in a city like Munich.