The other day, I was walking on the beach in Tokyo’s bay area. Comparing to other areas in Tokyo, I like its tranquil and peaceful atmosphere. However, suddenly I realized that the beach is full of plastic garbage. Some of them are large plastic chunks, and the others are pretty small or even tiny plastic fragments. But hey, who scattered the garbage all over the beach? It’s obviously us who did this. This is so-called “microplastics”, which pollute both the ocean and the land these days. Some of my friends have talked about microplastics before, but that day was the first time that I realized that the issue is pretty much real. This experience has been the main motivation that I decided to do something against this.
Because I am a software engineer, I thought I can make something that is based on deep neural network technology. It is more like a data-based approach rather than image processing, which is probably a better way because microplastics are very diverse in shapes and colors, which will definitely difficult for a conventional computer vision. The idea is that once the AI is ready, I will put it into a rover with a robotic arm, which autonomously removes plastics from the beach.
The key factor of this project is to build the AI that will tell plastics from mixed debris. More specifically, it will be an image classifier or an object detector using a neural network.
To do this, I need a whole bunch of training data. Luckily (or sadly), there’s a lot of it out there on the beach. So I grabbed some plastic debris and brought them back home.