Hello to the Mapillary forum,
After my initial disappointment with Mapillary and it’s sequence-oriented everything, I decided to try importing my Panoramio pictures again.
Unfortunately, most of the issues remain.
Compass angles are still messed up, and auto sequencing is wonky at best.
I tried looking into python scripts offered, however want to ask here first: are the scripts any better than what the web uploader does? Does sequence_split.py work better than the sequence bot on the web? I presume it would if you can set split distance (1km is too much) and time (does the web bot even check this?).
Also, why hasn’t anyone packaged export_panoramio.py into a web tool, or why doesn’t Mapillary offer a web GUI for it?
There are things they just don’t or cannot do and everyone is waiting for it for years by now.
2 Likes
I have never used the web uploader for anyting significant, but I know there are command line arguments for sequence_split.py so you can set both split time and distance.
The sequence oriented approch is clearly to mirror a street view service, but also to make it easier to use the images to train software that could run in self driving cars.
The lack of GUI is probably due to resources. They got venture capital but as much as a certain juicer, and developers are expensive.
Will you try to describe your issues in this thread and I will try to either help or say it is not possible?
yeah, issues/understaffing is understandable, although Mapillary has been around for a while.
I guess it is a better VC advert if you do something AI/ML/self-driving related, while some people migrated here thinking it would be a good Panoramio replacement (there isn’t one).
My issues all stem from this: exporting my photos from Panoramio was supposed to be automatic (?), but that never happened, so now you have to get the archive and go through the web uploader. The problem with it is its sequence cutter, which isn’t very good when all you are uploading is images taken at different spots, over a span of years, not minutes/seconds.
So I just wanted to know whether I should bother with the scripts or the web uploader has gotten better (clearly not). Time to get python installed again, I guess.
But seriously, packaging a py script for easier use shouldn’t be THAT hard. I mean, they wrote the uploader in the first place. Will have to wait couple more years for them to make it useful
I was capable of running python scripts. But I have a python versions mixup and I doubt that a proper user interface would help.
I have tried Anaconda and so on, but python is a curse.
I even went to python meetings for help.
Not a fan of python myself, but there isn’t any other option for Mapillary, the codebase is private
Regarding your archive, I assume you have your images in one or more folders and they are geotagged.
Using the Python scripts you can split your sequences. You can specify both cut time and distance, so if you have spend 30 min taking each image in a sequence of 10, you will get a sequence of 10. But the splitter will sort the images by date and time before splitting, so images taken in the same area on e.g. different days will not be put into the same sequence.
As you have written, Mapillary is sequence oriented, so stand alone images will often be put into a lot of single image sequences. That is also ok. Mapillarys algorithms will see that pictures are close together and add navigation between them, if they are close enough, no matter when they are taken.
I have never used the Panoramio importer, so I don’t know anything about it.
Regarding only making script, it usually more than doubles the development time just to add a gui. Mapillary have chosen to make a web GUI for people who needs that, but provides scripts for power users. That is a choice. Making a nice package of the scripts also takes a lot of time. Again that is a choice.
Regarding choice of programming language, there are various issues with all of them regarding distribution. Spending a lot of time on packaging, like some big commercial programs do, would fix that.
1 Like
oh, that’s what I’m looking for. Not sure whether the web uploader does that.
Looks like using the scripts is my best option, thanks for the explanations
As far as I know the webuploader does not split at all. I have heard of a job that will split images with more than e.g. 300 m apart, but that is not the real deal. Try the scripts and please ask questions.