This year the weather has been quite awful in Milan, and Sunday 20th it was raining once again. However, later during the day it was nice and even sunny, so I decided to go for another trip.

This time I planned in advance and went to Villaggio dei giornalisti, an area that was not particularly covered by previous footage. It is an area of architectural interest and particularly nice.

After that, I went back to the Parco Nord, covering more bike paths in the park.

This time I configured the camera correctly to record a time-lapse, getting around 4 GB of videos for a 2h trip, a much more reasonable file size. The battery also lasted longer, more than 2h and could have gone for much longer. I had to end my trip because it was getting dark.

Another thing I learned from the previous time was to use a fitness tracker to track my trip, so now I can see the calories and speed overall, and observe the trip later on Strava. I also could produce a GPX track, but they are already abundant on OSM in this area.

I still have to figure out an effective way to micromap based on these pictures, iD integrates with Mapillary but the interface is a bit cumbersome for large changes, currently I find easier to open another window with the footage only.

In addition to OSM, I occasionally take pictures of the areas I visit to upload them on Mapillary.

There are a few reasons I enjoy contributing to Mapillary:

• it constitutes an historical archive of pictures over time, seeing how a corner changed brings up memories and documents changes. I personally lived in two cities (Milan and Berlin) that changed a lot over the years and like to document it and observe these changes over time
• in Milan it can be used to document the space used by vehicles, including the irregular one, which is a detail not present in OSM tags
• it’s suitable for machine learning. I for one have a second job as a teacher for Data Science Retreat and a few students used this data to classify the quality of bike paths. Mapillary already automatically identifies, geolocate and segment common elements like street signs and manholes
• it’s suitable (sometimes) for 3D reconstructions for artistic purposes, like games (example here, it’s not based on ad-hoc images but mapillary could reasonably be used for it too. Mapillary calculates and provides point clouds too). I think this is an area where new approaches like gaussian splatting can do a lot
• it can, of course, be used for micromapping in OSM, these images show shop names, street etymology, fences and other details not easy to spot from satellite data and hard to add on the spot using StreetComplete or similar. Images from Mapillary are accessible directly in the iD editor.

For these and other reasons I contribute to Mapillary since years, using my phone with the Mapillary app whenever I visit a new or interesting place.

Imagine then my surprise when, a few months ago, I was contacted by Meta’s Reality Lab. They saw my contributions in Milan and offered me to borrow their GoPro MAX 360 which was not used anymore by another Mapillary contributor, to contribute as I see fit.

I had never used a 360 camera, or an action camera in general, so I’ll try to report here my experience in using it.

I want to thank Meta for this! Also to be clear I contribute to Mapillary and OSM and write this for fun and because I think it can be useful to others, I don’t work for Meta nor GoPro.

How does it work?

The camera is a 6cm square with two lenses, front and back, and a touch display. It has a mount for a bike helmet, and can be controlled via an app using bluetooth.

With this setup I can turn it on at the beginning of my bike trip and then just bike as usual. The camera has GPS, accelerometer and compass so a software can track and handle head movements and curves automatically, stabilizing the footage which is much needed when biking.

It stores the footage on an microSD card as a H.265 video with a telemetry track that is read by the Mapillary importer.

This is how a raw frame looks like when opened in VLC (you need the h265 codec of course):

After the upload the “seam” is reduced, and license plates and faces are blurred.

Where do I go?

For my first bike trip with the camera I went without a specific plan, not knowing how long it could eventually record, but followed a few criteria:

• I visited my “personal” places: my high school, where I usually go running, the area near my home
• I re-visited Viale Ca Granda, which got 360 coverage in 2020 from another user and where a few buildings went through a significant renovation to increase energy efficiency and can now be observed before and after
• I visited a park (Parco Nord), an area not particularly covered by Mapillary so far and on of the largest parks in Milan

How did it go?

If was fun!

The footage ended up taking a whopping 35 GB for a bit more than 1 hour. The battery ran off just before I arrived back home. I found out later that I had accidentally used the “video” settings, making a smooth video instead of a time-lapse with a frame every 0.5s (the recommended configuration for the bicycle). That’s why the videos files are so large (with the correct settings that would be around 2 GB and the battery lasts much longer). This extra data is essentially wasted since most consecutive frames end up having the same coordinates and are just ignored.

The fact I am recording drives me to try and visit unusual corners, places where I’ve never been even though they are just a few steps from home.

As a person who knows nothing about making videos I found fascinating how a 360 camera changes the approach to recording. You don’t have to point it to something specific like I did with the phone, you just know that everything you can see will be recorded (more, in fact). I can totally see why it’s used for outdoor “action” activities.

Compared to using the phone and walking this setup (bike + 360 camera) produces way more pictures and from a better perspective, the helmet mount is taller than the roof than most cars, which in Milan are parked pretty much everywhere, recording more useful details.

I noticed another user in Milan is uploading a ton of 360 images, by walking, and they seem to use a stick to keep the camera up, I guess for the same reason.

The camera is very conspicuous, and gets some attention, for example while I was at a park a ground of kids asked me whether I was a Youtuber :)

The quality of the pictures gets much worse at dusk, the ideal condition is during the full day but a bit cloudy to have indirect sunlight.

In addition to the mistake video format I realized afterwards that I had not tracked the trip with the phone. This is not a problem for Mapillary per se, since the GoPro has its own GPS receiver, but would have been nice to produce a GPX/Strava track too.

• Sparse OSM tools: https://www.openstreetmap.org/user/ivanbranco/diary/401437
• OSM Links from another user diary: https://www.openstreetmap.org/user/GrapeMapping/diary/401815
• Quick interactive analysis of the map stats and how they evolve over time (e.g. “how many trees are in the map in Italy”), includes estimation of completeness https://dashboard.ohsome.org/
• Gallery of cartography projects, mainly about Berlin https://hanshack.com/
• 30 Day Map challenge: https://30daymapchallenge.com/
• Even more beautiful maps: https://tjukanov.org/