Neural Network tuning campaign peek

So, here's a peek at an NN-tuning campaign I'm planning to start some time soon. I've got my Unity project to work with Google's Firebase storage and database.
What happens in the vid: In the first part I'm recording a facial expression each 0.4 seconds an then sending the raw image data directly to Firebase storage (and it's paths to Firebase database). In the second part it just downloads all the raw data and decodes it into a folder.
What I'm planning to do with this: when I tune the auth options of the database, I'll be asking people to help me train the NN - that is to make as many facial expressions as possible tied to a prior specified emotion. The machine will then try to guess each of them.
Each incorrect guess coming from the NN will then be uploaded to storage - and afterwards used as further training data for the machine! That way I'm hoping to sample the average player's expectation of what one or the other emotion should look like.
/ heathengarage

[@dmn4373]

Beautiful man.

[@gyr2365]

woah didn't know you can move his jaw