"Better than nothing" approach?


I’m currently the only active resource in a relatively large city (4M+ people), with fewer than 5 sequences when I started.
This means, of course, there is an enormous amount of streets that should be covered…

However, as most office workers, I usually drive around in the early morning or afternoon when the light conditions are not optimal: reflections from street signs, high contrast shadows, indirect sun darkening the picture (even when I see perfectly).

My question for you is this: should I consider an approach where having a less-than-optimal picture (for example, but not limited to [1]) is better than waiting weeks/months before I drive through that street in more appropriate conditions?

[1] http://imgur.com/wZdI8zM Here the light was perfect from my perspective, but the sun on one side affected the camera darkening the whole image.

What you can do is adjust the exposure in the app so that those darker areas are visible.

The downside of doing this is the sky becomes pure white in appearance, but hey, we all know what the sky looks like :wink:

1 Like

Hi @latin!

In the ideal Mapillary world conditions for mapping would be perfect but as we know reality is a little bit different. Therefore, we always say that it is better to take too many photos than too few as long as they are not blurry or black. If you happened to drive or walk on the same street when the conditions are better you can always take new photos and share them with Mapillary.

Happy mapping!

1 Like

a) I wouldn’t get hung up on getting every street in town. If there’s a neighborhood you really enjoy, use it as an excuse to get out and walk all of it and map it.

b) As for quality, there are a lot of different opinions in the community. Keep in mind that the pictures are data, not just pictures. There are already software programs that can use these so-so pictures to create nice 3-D models of the area, that is if they have enough coverage.

For myself, I use a Garmin Virb with a GoPro suction mount on the inside of my windshield. If it’s cloudy or getting dark I’ll tip it downward so it notices the lack of light.


As others have said or implied above I too (/four?) think that it’s definitely better to get more than less.

The example you gave is actually quite good. The only problem it really has is underexposure due to the light conditions. There’s been discussion elsewhere that those could also be post-processed in the Mapillary end in batch.

But even without that what I see is very clearly in the image is
(i) a three lane dual carriage (23 lanes)
(ii) asphalt paved road
(iii) without sidewalk
(iv) but with streetlights
(v) with a service road at least in the direction of the photo lane
(vi) with a (2
3) power line running on the right side (to the direction of the photo) of the road.
That’s a good chunk of information from a “bad” photo, a lot of which is or might not be visible in aerial imagery.

I am of the opinion (as an avid OpenStreetMapper) that if there’s even one piece of information that can be extracted from the photo – 3D models is an good example of something that one doesn’t necessarily come to think of – then it was worth it. And on top of that, when you take more photos and they are generally decent you simply don’t have the possibility to vet all of them. Yes, I try to remember bits that have been bad and then vet them before uploading but quite frankly not always. And I don’t fret about it.

With this “attitude” I’ve been “liberated” to take photos essentially when and where ever I drive (with some limits of course!) and am quite happy with what I’ve gotten covered, mainly in Nicaragua.

A quick example of a “bad” photo (series) of mine that I can think of is a recent trip to the mountain area. It was getting dark and already beyond what my Samsung Galaxy S5 (which is surprisingly bad in low light conditions) could handle. So I continued with so I continued with my Galaxy Tab 10 (which oddly enough performs better in these situations. But even with that there was just not enough light as there was just about 10 minutes to sunset. But I wanted to continue since especially because this stretch of the road had not been covered at all before. And now looking back (which I haven’t done before) I’m happy I did. See this example of a clearly-less-than-ideal photos. You can look around to South and North (or just hit play to see the sequence forward, towards the sunset and decreasing light) to see more examples of “bad photos”. But they all show at least the basics of the road, safety rails, etc.

Looking at this series I also spotted a new contributor who has also covered some previously uncovered territory here and clearly some of his photos are also “bad”. But they are covering virgin territory and are actually super valuable. See e.g. here and here for the “bad ones” (that all have some useful information) and then bit later up the hill how the photos very clearly show e.g. a transition from “normal” unpaved road to a narrow track type road (in OSM terms narrow=yes + perhaps tracktype=grade3 + surface=unpaved) in just three photos – and they also show a side road forking to the left in between. All of this is currently totally unmapped in OSM as you can see from the basemap. And on top of that, this sequence ends up right next to a gorgeous view point that overlooks Jinotega, perhaps the nicest city in the Nicaragua mountains. The last part of the stretch (starting from where the road turned into track type) it would be absolutely impossible to see from imagery you can drive up to the viewpoint. This would actually be of value for the tourism in Nicaragua. Not bad for a series of “bad photos”, eih?!


@jaakkoh, I agree, that sequence up the mountain is horrible yet it’s really good. That’s a great example that while not visually pleasing that sequence contains a lot of data.

1 Like