OpenStreetMap

My Bicycle setup for Mapillary

Posted by aharvey on 25 August 2017 in English (English)

I've started getting into capturing on Mapillary and have found it to be incredibly helpful for mapping on OSM. I wanted to go through my setup in case it's interesting for others.

bike setup

  • I prefer to capture 360 images
  • I use the LG360 (R105)
  • I bought two LG360 cameras off ebay, $125USD and $90USD. The LG360 is on the cheaper end of the scale compared to some of the others on the market.
  • Though there is a flaw in one of the cameras I bought which is a shame as you can see the left half of the image (front camera) is blurry.

LG360 lens issue

  • I'm happy with the image resolution and quality. I wouldn't want it any worse though. Most of the time street signs and parking signs aren't readable, but speed limits and destination signs are.

Destination sign

  • The LG360 is unreliable, sometimes it stops capturing, connecting to your phone can take several minutes and attempts before finally working, there can be gaps in the GPS tagging where it will use the last location for maybe 30 seconds (even though the GPS log might have data; in this case I end up using gpscorrelate, however I've found you have to play around with the time offset and even then you might get it right for the start of the trip but it will be out at the end).
  • I think the battery should last for around 1hr 45min at a 2 second capture interval.
  • I'm happy with the 2 second capture interval, but wouldn't want anything not supporting at least 2 seconds.
  • I'm using a 64GB Micro SD card XC I U1. Photos are about 5MB so you should get around 10,000 photos or 6 hours give or take.

  • I capture on my bicycle. I have a rack mounted which I attached an old broom stick I found on the side of the road to, which I then attach the Monopod to. It's all held together with Duct tape. The broom stick I'm using is light weight but does give some flex (if the camera had a built in gyro it could correct for it in software, I mean it's a 360 camera after all!).

  • I use this phone mount. It's great. Very easy to attach/detach your phone, has a Silicone band to hold you phone in place (I've had my phone come out of the claw once going over a pothole, and the band ensured it didn't come out). There is this ball joint which is stiff enough that you can just change the angle of your phone and it usually stays in place. I got mine for $2.10 USD.

bike phone mount

bike phone mount

pole setup

This pole is longer than the previous one. All the 360 photos on my Mapillary profile to date have been with the shorter pole. I'm still working out the best height to use. The higher the camera then the less I obstruct the view, but the more tree branches, signs etc I run into.

The main problem at the moment is it's a real pain uploading to Mapillary when you have something like 10GB a day and can only do around 90KB/second. I end up using rsync to a remote server as rsync doesn't mind being interrupted a hundred times and it's easy to limit the bandwidth. I also use lepton to compress the JPEGs further. WebP would be even better! The scripts I use for this are at https://github.com/andrewharvey/lg360-mapillary-helpers

If you get lucky on ebay parts are around $105 USD + duck tape, a bike, a phone and a Micro SD card.

Location: Victoria Cross, North Sydney, Sydney, North Sydney Council, New South Wales, 2060, Australia

Comment from SomeoneElse on 25 August 2017 at 09:27

I particularly like the "low bridge detector" :)

Hide this comment

Comment from aharvey on 25 August 2017 at 09:46

The other thing I forgot is, the LG360 needs to be connected to your moblie device while in use, which means the camera is wasting power with wifi and bluetooth. You can't just press a button to start capture and then again to stop.

There is no built in gyro so when it rotates your photos are rotate (so won't work well mounted on a helmet).

Hide this comment

Comment from ToniE on 25 August 2017 at 13:29

I have the same problems with my 1GBit/sec DSL upload speed.

I do the following:

  • prepare the images using mapillary Python script on my PC
  • rsync the images to my Raspberry-Pi
  • limit the upload speed to mapillary to 30KB/sec (by a traffic monitor on the Pi)
  • use the mapillary Python scripts on Pi to upload to mapillary
  • have a module for the RPI-Monitor to remotely monitor the progress

works fine for me. But for 5,000 images it might take 2-3 days to upload all images.

Hide this comment

Comment from utack on 2 September 2017 at 23:13

I like to capture using something with changeable battery.
Right now it is the GoPro Hero3, with 8 third party batteries that cost under 3€ a piece and run about 2h in 2s interval mode.
The camera mount by Tacx seemed appropriate, it combines the advatages of being affordable, being sturdy, holding the camera mostly steady when steering and being permanently attached to the bike without disturbing.
Tacx Gopro holder Imagemagick flips the result, fixes the extreme fisheye distortion, crops out (most of) the bicycle bags and the mapillary scripts sync it up with a gpx from my Garmin eTrex.
It has a narrow angle compared to your setup, but the advantage that it can capture for 12h+ with many batteries and does not need a phone.

Hide this comment

Comment from aharvey on 4 September 2017 at 01:52

@utack I'd like something standalone and reliable, but personally I find the benefits of 360 capture mean I'll sacrifice image quality and reliability of capture which it's easier to find in non 360 cameras.

I guess you'd need two, one for the front and one for the back?

Even without a swapable battery, something like http://www.ebay.com.au/itm/1-4-Expansion-Screw-Adapter-for-Ricoh-Theta-S-M15-LG-360-Cam-Samsung-Gear-360/262513821205 would raise the LG360 and let you run a USB cable to a power bank to keep it going for the whole day. You still need to power you phone though as unfortunately since the LG360 is so unreliable you need to keep the screen on as sometimes it just stops.

Hide this comment

Leave a comment

Parsed with Markdown

  • Headings

    # Heading
    ## Subheading

  • Unordered list

    * First item
    * Second item

  • Ordered list

    1. First item
    2. Second item

  • Link

    [Text](URL)
  • Image

    ![Alt text](URL)

Login to leave a comment