JOSM presets via voice control
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
osm.dic
osm.lm
osm_corpus.txt
readme.md
requirements.txt
sms-alert-4-daniel_simon.wav
speech2josm.py
tags.yaml

readme.md

speech2JOSM

Prototype of an external tool, that controls the Java OpenStreetMap editor by your voice. So you don't need to pick the most common OSM tagging presets manually and can speedup your tagging without interrupting your geometry editing.

speech2josm demo video

limitations

  • false positives detected esp. with noise
  • number of recognized tags is pretty limited
  • only english language
  • no numeric / ... custom values
  • JOSM control limited to tagging features only

setup

Start JOSM (>3850) and enable remote control for all actions (edit - preferences - remote)

  • sudo apt install pocketsphinx python-pocketsphinx pocketsphinx-en-us python-pyaudio python-requests (or use pip -R requirements.txt and python virtualenv for the python dependencies)
  • python speech2josm.py
  • wait and say 'footway' and confirm JOSM security dialog

tech

Contribute

  • extend osm_corpus.txt with control words
    • minimal command lenght, while beeing unique on all items
  • upload corpus to LM onlinetool and update osm.lm and osm.dic
  • add matching to tags.yaml and validate it