PlantTraits2024

Start Date: February 2 End Date: June 2 Competition URL: Kaggle

Description

This competition aims to predict plant properties - so-called plant traits - from citizen science plant photographs. Why are plant traits currently so relevant? Plant traits are plant properties that are used to describe how plants function and how they interact with the environment.  Due to global change, the environment and, hence, the biosphere are being transformed at an accelerating pace. However, we can hardly project on a global scale how plant traits and, as such, entire ecosystems will react to climate change because we do not have sufficient data on plant traits. A data treasure in this regard may be the very growing availability of citizen science photographs. There are already more than 20 million plant photographs available covering all ecosystem types and continents. The competition aims to predict the plant traits from the plant photographs using deep learning-based regression models (e.g., CNNs or Transformers). Any prediction that helps us better understand the global change on ecosystems is helpful, so please help us explore data treasure and the distribution of plant traits around the world!

Dataset

The dataset includes over 70,000 plant images from the iNaturalist project (obtained from Gbif). Each plant image combines ancillary information, environmental data (climate or soil properties), and satellite observations. The photographs and this ancillary information (x) shall be used to predict plant traits (y). Information on the plant traits was acquired from the TRY database.

Organizers

• Teja Kattenborn, [University of Freiburg, Germany]

• Ayushi Sharma,  [University of Freiburg, Germany]

Remote Sensing for Earth System Research (RSC4Earth, Leipzig University, Germany) https://rsc4earth.de/authors/tkattenborn/