Repo containing main patches for a system for classifying and broadcasting out EAO sound painting gestures. This repo DOES NOT contain the mediapipe skeletime app OR the skeleweb node app for heroku deployment. The main patches are:
- skeletime.BROADCAST - receives pose and multihand data from the mediapipe skeletime app and broadcasts that data via the skeleweb web app
- skeletime.RECEIVER - receives raw pose and multihand data from the skeleweb web app. Filters that data and sends it (locally via Max send/receives) to the skeletime.CLASSIFY patch
- /soundpainting-signs-gestures-recognition/Main_patch/skeletime.CLASSIFY - modified version of Arthur Parmetier's Sound Painting Recognition patch.
- navigate to the skeletime-mediapipe directory (on dispersion_noise this is ~/skeletime-mediapipe)
- run the command:
GLOG_logtostderr=1 bazel-bin/mediapipe/examples/desktop/skeletime_tracking/skeletime_tracking --calculator_graph_config_file=mediapipe/graphs/skeletime_tracking/skeletime_tracking.pbtxt
- if the application needs to be built, first run the command:
bazel build -c opt --define MEDIAPIPE_DISABLE_GPU=1 mediapipe/examples/desktop/skeletime_tracking:skeletime_tracking