iNat Challenge 2017

Overview

As part of the FGVC4 workshop at CVPR 2017 we are also conducting the iNat Challenge 2017 large scale species classification competition, sponsored by Google. It is estimated that the natural world contains several million species of plants and animals. Without expert knowledge, many of these species are extremely difficult to accurately classify due to their visual similarity. The goal of this competition is to push the state of the art in automatic image classification for real world data that features fine-grained categories, big class imbalances, and large numbers of classes.

The iNat Challenge 2017 dataset contains over 5,000 species, with a combined training and validation set of 675,000 images that have been collected and verified by multiple users from www.inaturalist.org. The dataset features many visually similar species, captured in a wide variety of situations, from all over the world.

Competition Details

Checkout the iNaturalist Competition Github repo for the specifics of the dataset and download links. The final results will be evaluated using a top-5 accuracy metric. The competition itself is being hosted on Kaggle.

Competition Results

The final rankings for the competition are as follows:

Acknowledgements

Thanks to the following people for their help organizing the dataset and competition:

• David Rolnick
• Weijun Wang
• Nathan Frey