It’s raining today, so no data capture :
Actually the time window for best photos is surprisingly short, just a few hours either side of noon.
I have been disappointed with the accuracy of the GPS coordinated gathered with my phone (Galaxy Note 8). It is too painful to drag each point back on to a background image, there are thousands to do. So it would be better to have them in the right place to start with. I am not in urban canyons, just urban area but quite hilly with a lot of vegetation.
The first idea I have had is to use SBAS (WAAS or EGNOS) which should do better and there is a trial service in my area (NZ). This needs a better GPS that a phone because it does not appear that phone gps software uses this extra facility. My Garmin 76c handheld does, and it also has provision for an external aerial so that may help a little. I would then have to synchronise and postprocess the images. This is a bit fiddly because I would still have to verify the locations. Can I do this in the web app? Maybe we need to write something in QGIS to a better experience.
The next option would be RTK gps. This is now possible with the Trimble Catalyst aerial that has a USB output that allows a direct connection to a phone. Would Mapillary consider supporting this? https://catalyst.trimble.com/ShopNow/
I know it all costs extra money, subscription etc, but if we are trying to build a street furniture database, How can it be accurate if the photos are not properly calibrated to the nearest metre or better?
Are there any other options to improve accuracy?