Competing in the Women in Data Science (WiDS) Datathon 2019
3 min read
Women in Data Science Datathon took off from the 29th day of January 2019 to the 27th day of February 2019. This was an online competition that saw the participation of 203 teams from all over the world. The aim of the challenge was to create a model that predicts the presence of oil palm plantations in satellite imagery. The partners of the Global WiDS team, Planet and Figure Eight provided us annotated dataset of satellite images taken by Planet satellites. The dataset images had a 3-meter special resolution and more importantly, each image was labeled 0 or 1 depending on whether an oil palm plantation appeared in it or not.
Some images of the Dataset used
Our task was to train a model that takes a satellite image as input and outputs a prediction of how likely it is that the image contains an oil palm plantation.
My team (Data Wranglers) was made up of two other members who happen to be the main contributors, Aseda Owusua Addai-Deseh and Kwadwo Agyapon-Ntra. We were given the liberty of using any framework for our data analysis. Initially, we used Pytorch and then Keras to build our model but we finally settled on Fast.ai. After training our model with the satellite images, performing validations, testing and fine-tuning the model several times we ended up with a prediction accuracy of up to 0.99526. This earned us the 110th position on the public leaderboard. The winning team had a prediction accuracy of up to 0.99957. Yeah, I know exactly what you are thinking, Wow! This competition was keenly contested and you can find more details here.
The journey through this one-month challenge taught me a lot, as WiDS gave us guidelines on how to go about the challenge and gave us resources to study with to hone our Data Science skills. WiDS introduced me to Kaggle.com which provides tutorials on Data Science related courses. This platform helped me a lot in my journey. Anything Data Science catches my interest now thanks to this competition.
My team had Abigail Mesrenyame Dogbe, the Lead for PyLadies Ghana, coaching, motivating and cheering us on the whole time. My team members were also very supportive and encouraging throughout the challenge. I had very limited skills in Data Science but they helped me study with weekly study outlines and materials.
Aseda is the Lead for the Data Science, AI, and ML group of PyLadies Ghana. She is a Data Scientist and a Software Engineer at minoHealth AI Labs in Accra.
She's passionate about using emerging technologies such as Data Science, AI and IoT to solve problems in society.
Dorothy is a final-year IT student at the University of Ghana Legon.
She has an interest in teaching others how to code and also manipulate data. More importantly, she has a passion for design, data manipulation, and Software Development.
Kwadwo is an Entrepreneur in Training at MEST. He is passionate about Africa and how technology and business can be used together to effect positive change on the continent, especially with the use of AI and Machine Learning.
You can find out more about Kwadwo here: https://kantra.xyz.
This competition was one of a kind and taught me a whole lot. If you are a Data Science enthusiast, try your hands-on opportunities like these and you will love how the challenges given will drive you to go beyond your limits and broaden your horizon in terms of skills and knowledge. There are many books and tutorials for beginners in Data Science. Get some of these resources and practice. I can assure you an experience you will not want to depart from.
Data Science is the real deal now!
This post was written by Dorothy Ewuah