Notes from Friday, September 14, 2018, micah@vrma.io wrote:
We created a deep learning plan called Pipeline.
Pipeline:
// create an event every time the raycast bounces off an object indicating head position,
// time log that event so we can correlate it with eeg data
// Time log every object in line of sight (object, angle, …)
We managed to create an event each time the ray casting is pointed at an object.
Next up that data needs to be time locked to the eeg data so we can export both sets of data to tensor flow
So we can train tensor flow on both sets of data
Then tensor flow will try to predict what we are looking at based on only eeg data.
Next up: Export all raw data in JSON at the end of web session to create training databases.
Train the data in tensorflow-gpu
Export the inference files to tensorflow.js for live inference in the browser
Explore also using tensorflow light and tensorflow raspberry pi
// R (Aria) setup a mapping from our local host to the noisebridge.net web page so folks can view the website at home from http://pegasus.noisebridge.net/webvr_temp/
Must ask R (via slack) to turn it on, if it’s off, because R has only granted us permission to have it up temporary (upon request for a specified time-interval)
Other notes:
We also tried to figure out why the websocket isn’t multi-threading by default, some of our research suggested that the we need to change a variable to true to enable the websocket to serve the EEG data to each instance of the client for each person who accesses our webvr page. We experimented with some different variations of websocket technology but we didn’t figure it out and ended up reverting to the original server script.
List of links from Neurotech SF VR on 8/31/2018
If you were there and you wanted any of the links that you saw on the screen tonight here is the full list.
Our Github where you can download a copy of the code we built together as a community. It’s open source:
Plasma Brain Dynamics Paper
Neural Lace Podcast S2 E1
Web VR code examples
https://aframe.io/examples/showcase/helloworld/
https://glitch.com/edit/#!/sparkling-talk?path=index.html:15:66
One path to integrating multiple sensor streams, time locking them, and presenting them to tensor flow.
other links you should check out are as follows
Main RSVP page for the event: https://www.meetup.com/NeuroTechSF/
Join our Discord group http://www.neurohaxor.com
Join the Global Neurotech Slack http://neurotechx.herokuapp.com
Make sure to join the SF channel: #_san-francisco
Join this Facebook Group: Self Aware Networks: Computational Biology: Neural Lace https://www.selfawarenetworks.com
Join this Facebook Group: Neurophysics+ https://www.facebook.com/groups/IFLNeuro/
Join this Facebook Group: NeurotechSF https://www.facebook.com/groups/neurosf/
Join this Facebook Group: Neurohaxor https://www.facebook.com/groups/neurohaxor/
My business card is at http://vrma.work
