So it’s taken a month to get back on (ended up having to re-route my email to a server that didn’t bin the password reset messages), but I’m uploading again. My upload backlog has grown to 750k photos in that time
Some random questions about old sequences, and the new upload process:
From watching the new mapillary_tools at work, I see that there’s an extra step been added to uploading - the collection of files are compressed into a single temporary archive file and that’s what gets uploaded. I generate a lot of photos but only have a slow upload connection, so my batch uploader runs two threads - one getting sequences ready to upload, one uploading those prepared sequences. This runs on a Raspberry Pi accessing the photos on a NAS. The prepare stage is slow, but this doesn’t matter as it’s not the bottleneck. However, the new compress stage is part of the ‘upload’ step - so for every sequence of 500 images the uploads stop for 20 minutes while it creates a 3GB temporary file on the SD card, so the scarce resource - upload bandwidth - now lies idle for 10-20% of the time. Question #2: could the compress stage be moved to the ‘process’ step, so the ‘upload’ step can just get on with uploading?
I’m not currently using the phone app for capturing, but when I did I often filled the phone’s storage to the brim, sometimes with just a single, huge sequence. Question #3: out of curiosity, what happens with the apps now if the storage is full - presumably it won’t have space to create the archive file to upload?
Sequences being uploaded immediately before the changeover:
Question #4: I had a batch uploader running using the previous version of mapillary_tools at the time the changeover happened. A fairly random selection of 32 sequences it uploaded have not appeared on the map or my sequences view after a month - will these eventually appear, or do I re-upload?