Global #CompletetheMap III


#41

Good feedback and definitely things for us to factor into the next CtM challenge. Let me respond to the points one by one.

  1. The score is not necessarily being multiplied by two. The multiplication depends on the number of people you have contributing. You’d need 10 people in the area for this to happen which is an impressive community achievement. The idea is to counter the issue that you mentioned, encouraging people to participate in areas where Mapillary/mapping is less popular. This is system is definitely less than perfect however, and vulnerable to manipulation. We’ll keep an eye on it and review for CtM IV!

  2. I’m not sure this is a primary concern. If people decide to contribute to one account, they also have to share the prize. Interested to hear why this is a concern.

  3. This is a really important addition and would hopefully incentivise honest and effective image capture. Stay tuned on this front, things are in the works :wink:

  4. A new challenge could reduce the emphasis on UKM. Keen to discuss next time we meet to find out where we could find a balance.

  5. Very validate point. The challenge doesn’t balance this very well at the moment, but it is unlikely we’ll have a solution to this anytime soon. That being said, @soren_johannessen and @nunocaldeira were both walking considerably in their locations in the last challenge and both emerged as winners. They were creative and found new parts of their city that needed mapping, many of which were off the main roads.


#42

@westbam UKM is calculated like this:

  1. map match (align) sequences (green lines) to road segments, so we know which sequences map which road segments
  2. and then we count the length of road segments that you firstly mapped as UKM

For example, assume we have a road segment of length 10KM, three sequences mapped this road segment in this order:

  1. A firstly mapped 0 -> 5KM (first half)
  2. B mapped 3KM -> 7KM
  3. C mapped 5KM -> 9KM

Then we get

  • A’s UKM is 5KM
  • B’s UKM is 2KM
  • C’s UKM is 2KM

The total UKM of the 3 sequences is 9KM.


#43

Is each direction of a road segment considered unique, or does one direction complete a road?

And what kind of roads are included in calculation? Just actual roads, or would it include highway=service like alleys and parking lots?

Also, I assume the UKM precludes any roads mapped before August, right? So covering those lines again wouldn’t count as UKM?


#44

Direction doesn’t considered. Mapping in one direction doesn’t count if the road is mapped in the other direction.

All navigatable roads are included. Roads are generated with https://github.com/valhalla/valhalla

If the CtM session is for August, if you map a road that was mapped before August it’s still count as UKM.


#45

Sorry, I am not really understanding. The city center in my home town was already well photographed. If I go those roads again now in August, will those count as UKM?

Also, if I cycle on a cycling path that is parallel to a road that is already covered (and on OpenStreetMap it is drawn in as a separate path), does that count as UKM?

Taking photos from a tram, train, river boat, do those count as UKM, or is it only actual roads that count?


#46

User teamwork is managing to take 320 photos a minute! :dizzy_face: That means they’re scoring a point for every 3 minutes on the road, not even counting the unique kilometers.


#47

Suggestion for the next Complete the Map: URM – Unique Road Minutes. Instead of counting kilometers and photos, count minutes. Counting kilometers gives car-owners an unfair advantage, and counting photos gives high speed cameras and videorecorders an unfair advantage. If you count minutes, you get a much better reflection of how much time a user if willing to invest.

To prevent people just standing still while making photos (which is currently not covered either) you can cut sequences (for scoring purposes, doesn’t have to be cut on the map) whenever there is less than 10 meters movement within one minute, and discard all those sequences.

(Of course, counting time could be cheated by editing the timestamps of the photos, stretching a 10 minute sequence to several hours…)


#48

This team is working on an ambulance because they get so quickly :slight_smile:


#49

I think that’s actually a great idea. If Teamwork put in the time I have, he’d have something like 4 times the points… I’m busting my ass just to keep up.


#51

What is happening here? I’ve uploaded 10.000 photos from a 20km cycling trip, taking care to cover as much as possible areas that I haven’t covered before (not even before august). Yet, my UKM has not changed!

(Several other users did not increase their UKM either, but they didn’t make many photos.)


#52

Do I need to move my path closer to the center of the road?
For the system to count the UKM.
0ec2c-clip-38kb


#53

I totally agree that the scoring system is unfair right now, I shouldn’t get so much score for the number of images, just because I have a dashcam, but I also disagree with the weight of UKM because you have no chance if you live in an area that is already mapped. (Feels even worse if you are the one who mapped it.)

They should also reward if you take pictures that are 3 months newer than the previous ones or something like that - for less points of course. A 360 photo should worth twice as much, since that covers both directions. And the multiplier for participants is a complete joke. It doesn’t encourage anything else than cheating.

Sorry for my grammar.


#54

@teamwork, if I understand right, UKM does not look at what has been mapped before august. It’s about unique within this competition. But I am not 100% certain, the information is not very clear. Anyway, the west of Budapest is still quite unmapped. You can make a lot of UKMs there, regardless of how it is counted.

For your information, I am also the one who has already mapped all the streets in my area. In addition, I don’t have a car. So all my kilometers are done by bike, foot and public transport.


#55

On choosing a place to nominate:

I even considered nominating a city in my neighbouring country that is barely mapped and seems more dense with roads, tho the (square) grid would still end up with lots of wooded areas because of the shape of the city. I ended up picking my center grid about 20km from home after carefully drawing a 7kmx7km square in JOSM. 1st August was rain day and my grid was not even ready yet, so I skipped. 2nd August turned out to be my only photomapping day as it was raining much of the first week. By then I decided to quit when I realise there is no way I am getting top 3 no matter how much petrol I upload into my scooter. My country is just not dense enough. I figured it is better to do my mapping at my own pace.

PS: I thought at the start that if I only needed to do like 1000km on my scooter, that would only cost me about USD7.30 on petrol, and I have access to unlimited and decent-speed internet. So, how much are you people spending for this??


#56

Some good feedback on the scoring system.

I agree changes are needed for the next round of Global CtM. All of this will be taken on board, but keep in mind the technical limitations that make it harder to implement some of these scoring systems.
Unique road minutes could be a helpful variable to include, but it poses its own questions. What is the maximum acceptable time between two images being taken? If someone captures 1 image very minute, are their contributions as valuable as someone like @teamwork taking multiple images a second?

I think the main tweak needed right now is to the participant variable. It’s quite heavily weighted at the moment and does not factor in how much that individual contributed. The idea is pretty simple though, it encourages the participant to get other people in their area involved.


#57

I guess the value depends on what your goal is. When updating OpenStreetMap, I have had lots of help from Mapillary sequences that consist of single images taken here and there, sometimes more than 5 minutes apart. In that case, there is hardly any (added) value to taking 320 photos a minute. However, if the goal is to feed AI and SfM routines, then that overload of images may be very valuable. When taking frames from video,
you can just as well go up to 1800 photos (or even 3600 or 7200 photos for HFR video) per minute. Then you’d have 500 points in two-four hours.

Sequences like https://www.mapillary.com/map/im/5bDYCgiKFuxu9HJM1KYvPg and https://www.mapillary.com/map/im/WH_DHcTV12p1AKlki8-wig are great valuable examples. For me, AI and SfM are no more than a nice by-product.

But I agree, you need a maximum distance. Taking one photo today and another tomorrow in the same sequence does not mean I’ve been mapping continuously for 24 hours. However, that should be relatively easy to automate. For OSM usage, 5 minute intervals could still be considered a very useful sequence.

I think the participants variable should be a one-off bonus, not a multiplier. It’s a bit weird that an active mapper can double their score by having 9 other people take 1-2 photos.


#58

In the start, I’ve taken a few bus trips, payed maybe 12-15 euros for that. After that, I’ve only been bicycling and walking, which costs me no more than some fruit and nuts. However, I see now that my bike is ripe for a maintenance, which could cost a bit…


#59

I’ve exclusively biked for this, and the increased mileage sure adds fuel costs, but who minds eating more? I have punctured two tubes, though, because the roads around here are pretty awful, so add some bucks there. And I probably use a good bit of electricity processing and uploading the photos.


#60

@eneerhut, @tao could you please clarify for us unambiguously, if I cover a road that has already been covered before August, will it or will it not count as UKM?


#61

Yes, it will still count. Sorry for being not clear here.