CSIRO - Image2Biomass Prediction

Overview

This hackathon challenges participants to build models that predict pasture biomass using images, ground-truth measurements, and publicly available datasets. The goal is to help farmers make smarter grazing choices, improve pasture health monitoring, and drive the agriculture industry towards more sustainable practices. You'll work with a dataset of Australian pastures across different seasons and regions.

Requirements

Participants need to build a model that predicts pasture biomass. The submission must be a CSV file in long format with two columns: 'sampleid' and 'target'. The 'target' column should contain your predicted biomass value (in grams) for specific target names like 'DryGreeng', 'DryDeadg', 'DryCloverg', 'GDMg', and 'DryTotalg'. Submissions must be made through Kaggle Notebooks, with specific runtime limits for CPU and GPU, internet access disabled, and the use of freely available external data allowed. The submission file must be named 'submission.csv'.

Prizes

There is a total prize pool of $75,000. The first-place winner will receive $50,000, the second-place winner will get $20,000, and the third-place winner will be awarded $5,000. Participants can also earn awards points and medals.