Is a neural network better than Ash at detecting Team Rocket? If so, how?


Go to NumFOCUS academy page.

In the world of Pokemon, there is a running gag in the inability of protagonist Ash Ketchum, at recognizing the series antagonist, Team Rocket. This fact got me thinking: is a neural network better than Ash? To answer this, I’m using TensorFlow, TensorFlow.js, and Google AutoML to visualize a network’s activation maps to interpret why it says this is Team Rocket.


Juan De Dios Santos

Juan is a Data Engineer at LOVOO, a dating, and social platform. As part of the LVPD team, he’s responsible for developing methods to detect and combat the fraudsters and spammers encountered in the platform.

In his free time—when he’s not dealing with fraud—you will surely find him experimenting with personal data to learn curious details about himself or having fun with object detection models. You can find more about his work at

Juan holds a BSc in Computer Science from the University of Puerto Rico - Rio Piedras Campus and an MSc in Computer Science from Uppsala University in Sweden.