Model Card for CetaceaNet

We provide a model for classifying whale species from images of their tails and fins.

Model Details

The model takes as input a natural image of a cetacean and returns the three most probable cetacean species identified in this image.

Model Description

This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.

Developed by: HappyWhale
Shared by [optional]: The Saving-Willy organization
Model type: EfficientNet

Model Sources

Repository: https://github.com/knshnb/kaggle-happywhale-1st-place
Paper: https://besjournals.onlinelibrary.wiley.com/doi/10.1111/2041-210X.14167

Uses

This model is intended for research use cases. It is intended to be fine-tuned on new data gathered by research institutions around the World.

Downstream Use

We think that an interesting downstream use case would be identifying whale IDs based on our model (and future extensions of it).

Out-of-Scope Use

This model is not intended to facilitate marine tourism or the exploitation of cetaceans in the wild and marine wildlife.

How to Get Started with the Model

Install the necessary libraries to run our model (transformers and the extra requirements.txt):

pip install requirements.txt

Use the code below to get started with the model.

import cv2
from transformers import AutoModelForImageClassification

cetacean_classifier = AutoModelForImageClassification.from_pretrained("Saving-Willy/cetacean-classifier", trust_remote_code=True)
img = cv2.imread("tail.jpg")
predictions = cetacean_classifier(img)

Training and Evaluation Details

To learn more about how the model was trained and evaluated, see 1st Place Solution of Kaggle Happywhale Competition.

Citation

If you use this model in your research, please cite:

the original model authors:

@article{patton2023deep,
  title={A deep learning approach to photo--identification demonstrates high performance on two dozen cetacean species},
  author={Patton, Philip T and Cheeseman, Ted and Abe, Kenshin and Yamaguchi, Taiki and Reade, Walter and Southerland, Ken and Howard, Addison and Oleson, Erin M and Allen, Jason B and Ashe, Erin and others},
  journal={Methods in ecology and evolution},
  volume={14},
  number={10},
  pages={2611--2625},
  year={2023},
  publisher={Wiley Online Library}
}

the HappyWhale project:

@misc{happy-whale-and-dolphin,
    author = {Ted Cheeseman and Ken Southerland and Walter Reade and Addison Howard},
    title = {Happywhale - Whale and Dolphin Identification},
    year = {2022},
    howpublished = {\url{https://kaggle.com/competitions/happy-whale-and-dolphin}},
    note = {Kaggle}
}

Saving-Willy
/

cetacean-classifier