At Figure Eight, we’ve always been fans of the Women in Data Science Conference (WiDS). They inspire people from across the world to hone their data science skills, they focus on underserved populations, and they promote data for the common good. Based in nearby Stanford, WiDS now reaches tens of thousands of (virtual) attendees and is growing every year. It’s a tremendous conference with a good heart and some great content and lessons for data scientists at every skill level.
All of this is to say that when the opportunity to team with WiDS came along, we couldn’t have been happier to help.
A little background: one of the yearly events WiDS hosts is a datathon. If you’re unfamiliar with the term, think “hackathon, but with data.” In their words, “The WiDS Datathon aims to inspire women worldwide to learn more about data science, and to create a supportive environment for women to connect with others in their community who share their interests.” Crucially, the WiDS datathon is meant to be both accessible to novices and experts while dealing with a problem of global importance. This year’s challenge focuses on oil palm plantations and the data that’s powering the datathon comes annotated by your friends at Figure Eight.
Before we get into the specifics, let’s talk oil palm. Oil palm plantations are a gigantic environmental problem, chiefly because of the attendant deforestation and displacement concerns. These plantations are multiplying, destroying biodiversity in dozens of countries worldwide, and currently cover 27 million hectares of the globe. Why is that? Well, palm oil (the byproduct of oil palms) is in half of all supermarket products. The demand is intense. But so are the consequences of the plantations.
Understanding where these plantations are and how rapidly they’re growing is a big challenge. And that’s what the 2019 WiDS datathon is about. In fact, if you’d like to sign up and participate, now is a good time to provide that link for you.
So how exactly did we help? By annotating data, of course. The Oil Palm Datathon is made up of thousands of satellite images, but those images were raw and unlabeled. These came originally from another WiDS partner, Planet Labs, and our role was something we know: image classification. Because the WiDS datathon is aimed both at expert and novice data scientists, it was important that we do simple annotations (as opposed to pixel labeling, for example), so the labels here are binary: “is there oil palm in this image?” with just “yes,” “no,” and “unclear” as options.
We were but a small part of the WiDS datathon but we’re really proud to be included. They do great work and were enthusiastic partners throughout the process. If you’d like more information on WiDS (or want to attend), please head to their website. The datathon is up and running as we speak and, if you’d like a little background before you get going, they’re running a webinar this Thursday to get you started.