Intro to the Kaggle platform for scientific research with CSIRO's Image2Biomass competition
Join the GDG AI for Science for a hands-on workshop designed specifically for researchers, scientists, and students who want to learn how to leverage the Kaggle platform for data-driven discovery. Kaggle is a rich ecosystem for accessing highly-curated, domain-specific datasets, collaborating in notebooks, leveraging free cloud GPU resources, and, as a bonus, offers competitions to test your workflows and skills and win cash prizes.
This workshop will demystify the process of participating in a data science competition and catalyse the Kaggle ecosystem for your research. We will work with CSIRO's live competition Image2Biomass, tackling a real-world scientific imaging challenge to help farmers estimate feed for livestock.
Expect a practical, end-to-end walkthrough as we work directly within Kaggle Notebooks to build, train, and submit a solution. This session is designed to give you a complete framework you can apply to future competitions and your own research work.
Join Nathaniel Butterworth (Google) and Qiyu Liao (Data61/CSIRO) to help you gain an understanding of:
Navigating the Kaggle Ecosystem: You'll learn how to effectively find, fork, and manage Kaggle notebooks and datasets for your own research.
Modern AI for Image Segmentation: You'll learn fundamental AI/ML approaches in Python for image-based prediction.
The End-to-End Competition Workflow: You'll get practical experience with the entire process, from exploring the data and building a baseline model to generating a valid submission file.
Making Your First Submission: You'll gain the confidence to officially join the competition, submit your results to the live leaderboard, and understand how to iterate and improve your score.
00:00:00 Overview of Kaggle
00:11:04 CSIRO Image2Biomass Prediction competition
00:33:42 Code your first competition Notebook
01:02:06 Advanced Computer Vision Notebook