Machine Learning

For all the toil that went into developing the pipeline and deep learning model for my ML baby monitor project, I discovered that my model didn't actually solve the problem at hand. This kind of gap exists in the real world of ML development as well, and there's a critical insight from the field of User-Centered Design that could be the key to bridging it.

Over a year ago, I wrote a couple of blog posts about how I trained a TensorFlow model on audio clips of my crying baby so I could get some insights into his sleep patterns and behavior. I distilled this experience into a PyGotham talk titled "Can Neural Networks Make Me a Better Parent?" and it reflects some extra maturity in my machine learning knowledge, as well as some hard-won parenting wisdom and experience.

In which we train an TensorFlow model with the tears of my little one, then deploy it on a Raspberry Pi.

How do you deal with the travails of colic and new parenthood? You plunge yourself into an ML project, of course. Here, we discuss a naive solution to detecting baby cries via an onboard microphone on a Raspberry Pi.