inat_comp

Why we included this project

The iNaturalist competition datasets have become a common benchmark for species-level recognition, and this repository gathers them in one place. Each year's folder holds the dataset description, class taxonomy, evaluation protocol, and download links from the FGVC workshops at CVPR, so you can grab a specific year's benchmark without digging through workshop pages. It's reference material rather than runnable code: it explains what the data contains, how the train/val/test splits are organized, and how submissions were scored. That makes it useful for researchers and students who want to reproduce published results or test a new model against a widely cited dataset. Teams planning a biodiversity or species-identification product can also get a sense of the scale and difficulty before collecting their own data.