Wandrer for 360 Maps

There is a Strava-based website, Wandrer - https://wandrer.earth .

It marks roads as cycled, awards points for the most roads cycled in some region and allows to download maps/routes for roads not covered yet.

Imagine a tool that…

  • Shows roads (and paths!) without [recent] imagery.
  • Awards points for covering s stretch without images, or without recent images.
  • Non-covered sections get the most points. Sections with existing images gradually accumulate points, as the imagery gets older.
  • Points are based on image type/direction (360 vs directional, one direction vs another, maybe image angle etc) and frequency (distance).
  • Perhaps eventually even scoring images based on sharpness vs blurriness…
  • Routes can be generated based on various sliders (to prefer roads without any images vs roads with older images and so on) - also based on road category (for example, motorcyclists might want to exclude paths and/or unpaved roads etc).
  • Routes can be set “A to B with n extra kilometers to collect the most points”, where A and B can be the same spot.
  • Routes can be downloaded in various formats to use them in GPS devices/phones.

Is anyone working on something like this?

I love this!

What a cool app.

The scoring system you describe is very similar to Mapillary’s Complete the Map: http://www.completethemap.org/

We’re actually working on similar gamification of imagery capture in Map the Paths.

Here’s a little teaser:

Seems there is a good opportunity to team up with Wanderer on this too. I’m going to email them.

Questions, questions… :wink:

Is this system going to be “360 only”?

“leader board” based on distance mapped? I would like to add the suggestion to add scoring as to if someone has mapped that route previously… At mapillary the leader board is only the amount of images taken… one places a camera in his car and uploads his every day route to work… Nice for the numbers, absolute horror for database quality!!

Please add something like: if image was never taken, or more then five years ago (outside of the current trace) within 10 meters (maybe double with routes without paths and half in cities?): allow 100% points, if it was taken a year ago 50%, two years 65%, three 80%, four years 90%. half a year: 25%, three months: 10%, one month 5%, less then a month: 1%.

That way award people adding NEW traces and discourage getting more of the same…

1 Like

Now that I’m on a roll… :wink:

Don’t use distance mapped, use images taken. In my opinion the perfect distance is 3 meters between images, on a highway much more (double or triple that, lett’s say 10 meters)). But in the context where we are talking I think you can fix it (for now) to 3 meters.

Thus (going into percentages again) on or above 3 meters: 100% score
2 meters: 71% (2/3 = 66%+5; give slightly more then at 3 meters)
1.5 meters: 55% (1.5/3=50%+5, just slightly more then 50%…)
< 1.5 meters I’dd say remove half of the images from the process!! (effectively going to 3 meters…)

(formula is a quicky, prevent from getting more at 2.99 meters then at 3 meters… but you get the point I hope)
If you find this a bit harsh you could go to 1 meters and then go to doubling (effectively going to a minimum of 2 meters)

Mapillary (my main reference in this) went mostly for quantity and attempts to increase quality because of that… When “we” are going “full 360” (i hope?) we do not need to get to a system where multiple “flat images” get combined into a 360-like image (fantastic piece of programming… but never as good as a single good 360)…

I say go for quality, in the broader sense of the word. And that goal should be reflected in the scoring model…

I wrote this @mapillary last year:

do not underestimate the power of scoring / ranking.
In “the good old days” I was one of the top Dutch contributors of the Berkeley SETI@home project (distributed computing project to find ET). The only thing gained were points and a team one could join. I built the statistical engine and graphics for the top Dutch SETI-team. We had competitions with other teams, other members etc! Why? Because we loved the concept and… well never underestimate the power of scoring / ranking :stuck_out_tongue:

Above all, the scoring system must be transparent and fair… I take great pride in that 100% of my uploaded images are 360 degree images. That doesn’t give me anything in the “leader boards”, I don’t mind, but I do think it would help if Mapillary would implement a scoring system that would be more then just 1 image = 1 point. Give more points to better images… like in that complete the map link earlier (and do give +2 for 360 images :stuck_out_tongue: ), give like two extra points for never mapped roads, one extra point when the previous image of that spot is over two years old, 1 or so points for an image that has been made less then a week ago etc… something like that!
I realize that would cost some extra CPU power for Mapillary, but I am sure the overall image quality would improve… in the end, quality must win over quantity, must it not?

AD: I think a scoring system I wrote about earlier is even better :wink:

PS: to add to the philosophical discussion that started after my post. There is also quality in quantity. But I dare to write there is even more quality when there is more quality in the quantity.

@Eesger some great ideas.

The problem with generic leader boards is they are not focused on a use-case.

For example, for us here at Trek View, seeing the same location on different days is actually very valuable in training our computer vision software. For people looking at the images, this is boring.

Speaking to some of the Mapillary team, their advice was to keep it as simple as possible.

I believe a formula weighted on imagery (new vs. old, 2d vs. 360, resolution) and transport method (clean vs. fossil fuel) is a good start, fairly generic and lends well to being transparent, as you suggest.

Secondly, the leader board could always be filtered for more specific use-case (e.g. only show me images captured on a bike, or 360 images). Prizes could be awarded in this way.

What do you think?

Ah! Well, that is a goal you hadn’t told me about :wink:
I don’t have a good picture as to what your company & goal entails… but if your goal is (like Mapillary) to “simply get as much images as possible” for vision training purposes, well, yea…

Well, that was a part where I (strongly) disagree… but that is also very depending on the goal…

The scoring at Mapillary was quite obvious of their goal… “plus 1 for each image”; thus: we simply want as much geo referenced imagery as possible.

I am much more thinking along the line of an “open sourced” streetview, thus I value 360 imagery much higher (I personally would go for only 360 :wink: ) and with this in mind there is no value in the same route covered again in a few weeks or months (assuming the “first track” is of good quality)

So whatever scoring system you come up with: it should embody your goal!

HQ | Blog | Spotted a Trekker? | Become a Trekker | Facebook | Instagram