Local-NLP-Backend
Local-NLP-Backend copied to clipboard
Helium314
[Feature] Manual training
I wanted to ask if manual training could be something you could consider to add.
It happens that I handle devices which very low GPS detection and in zones were GPS services are not available.
Since this work attaching a GPS location to a phone cell of WiFi AP I was thinking about the possibility of doing this manually. Perhaps, maybe, through a map choosing your current location and with access to a table or log to remove it in the case I was wrong at any time.
After adding manual locations, GPS detection could be tried first and if it fails, rely on manual overrides for zones.
Using a map would require the backend to acquire the internet permission, which is against the basics of this app.
Adding data is already possible using the import functionality, but this requires preparation of the data some other way and is not very convenient. Maybe there is some way where only entering coordinates is required...
Downloading maps for online use is an option?
Maybe similar to the way SatStat does it.
Manual coordinates sounds counterproductive from mobile when you just want to add quick datapoints of your location.
Using on-device mapsforge maps should be possible, but this looks like it's considerable work. I will not do this (anytime soon), but if someone wants to contribute the necessary code it could be implemented.
A different way would be handling geo uris, see https://developer.android.com/guide/components/intents-common#ViewMap This should be relatively easy to implement, but you will need a maps app that allows you to share locations using this format (e.g. OSMAnd).
Is it possible to create a bounty in order to implement this?
I'm currently not interested in setting this up (and even reading on how to do it properly). However, I did implement to geo uri handling in 1.2.9. You can use it e.g. with OSMAnd by choosing share -> "geo:" and open with LocalNLP.