Using Bigfoot to explain training ML models with synthetic data

Bigfoot, also widely known as Sasquatch, is a large, hairy, ape-like creature that is said to inhabit the wild and forested areas of North America. It’s rumored to be 6 to 15 feet tall, and muscular, feasting on deer, elk, and bear, among others¹².

Naturally, you wouldn’t want your child to get close to one in the wild.

Bigfoot in a forest, viewed from the distance — This and all images of this post were generated with Google’s Nano Banana image model.

So, during playtime, you try to explain what Bigfoot is and why it’s dangerous.

A cartoon of a father desperately trying to explain to his toddler son what a Bigfoot is.

But your child doesn’t quite understand, so you use clay to model one. Hopefully your child got the idea now.

The father uses some Playdough to model a bigfoot

Moving from this analogy to the machine learning domain, you have trained a perception system to detect a Bigfoot solely by providing it with synthetic data about what is a Bigfoot and what is not a Bigfoot.

Now you hope that this perception system can correctly detect a Bigfoot in the wild.

The child sees the real Bigfoot and is startled

https://en.wikipedia.org/wiki/Bigfoot ↩︎
https://worldbuilding.stackexchange.com/questions/261606/what-would-the-natural-diet-of-bigfoot-be ↩︎

Using Bigfoot to explain training ML models with synthetic data

More posts

Using Bigfoot to explain training ML models with synthetic data