For this competition, we have curated data from the broader [FathomNet Database](https://database.fathomnet.org/fathomnet/#/) image set to tackle hierarchical classification. The training set contains 79 categories of marine animals of varying taxonomic ranks (e.g., family, genus, species), wherein 300 example instances of each category are provided for training. The test set contains the same 79 categories, where roughly 10 example instances of each category are provided for evaluation. The challenge is to develop a model that can accurately classify these taxa and ideally leverage their taxonomic information to do so. Developing these solutions for ocean research will enable scientists to process and explore ocean data more efficiently
Participants should not use other image sets to train their models outside of the Kaggle competition data and the FathomNet Database [website](https://database.fathomnet.org/fathomnet/#/), however they can leverage other types of outside data sources for other aspects of the competition. Pre-trained models can be used for initialization of training (e.g. [BioCLIP](https://imageomics.github.io/bioclip/), [ImageNet](https://www.image-net.org/), or [COCO](https://cocodataset.org/#home) classification or detection models provided by many deep learning packages). Teams should plan on specifying additional data sources and/or pre-trained models used when uploading results.