Positivitypunk Backpack

The Positivity Pack's machine learning gesture recognition function is fully operational!

The model is a CNN built in Python with TensorFlow, trained on acceleration data. Gestures are recognized and sent over to Twitch.tv for my live streams!

New Code Improvements are:

- Socket is non-blocking

- Machine learning Gesture recognition added - see project for more details on that end.

Updated code is in the github

As I am still building aspects of the code for filtering and NLP of the messages, I still need a quick way of displaying the messages already submitted by everyone.

Let's work with Davinici Resolve, a video editing software that is free for the base version.

I created a timeline and generated a solid white background from the background generator.

Then I added a second video layer on top of the white background. Adding a Title to that.

With the title selected in the timeline change the text and font to taste.
Add more titles for each of your positive messages. Then render the entire video as an mp4


Transfer that mp4 file over to your raspberry pi in the pack. Play it back with VLC player, using the fullscreen mode. It'll look something like this!

Next steps.

Use a python script to swap between displaying images of the text or use it to swap between videos using the VLC python keybindings in the python-vlc library.

Then setup a an irc connection to Twitch chat and let them control the messages displayed on the backpack.