Iterative Human and Automated Identification of Wildlife Images

[Paper] [Preprint] [Press]

*This repository is the author's re-implementation of the iterative recognition system with machine and humans in the loop described in:

"Iterative Human and Automated Identification of Wildlife Images" Zhongqi Miao*, Ziwei Liu*, Kaitlyn M. Gaynor, Meredith S. Palmer, Stella X. Yu, Wayne M, Getz in Nature - Machine Intelligence, 2021*

Further information please contact Zhongqi Miao and Ziwei Liu.

Requirements

• PyTorch (version >= 0.4.1)
• scikit-learn

Data

All raw camera trap images that were used in this study (except classes with humans), along with the associated annotation information, are uploaded to the publicly-available Labeled Information Library of Alexandria: Biology and Conservation (LILA BC), and can be downloaded [here].

Changing dataset root for training and testing

Once the data is downloaded, please change the data root in the configuration files. For example: dataset_root: /Mozambique.

Stage 1: pre-training and evaluation

python main.py --config ./configs/Stage_1/plain_resnet_MOZ_S1_101920.yaml

Stage 1: energy fine-tuning

python main.py --config ./configs/Stage_1/energy_resnet_MOZ_S1_101920.yaml --energy_ft

Stage 2: training

python main.py --config ./configs/Stage_2/pslabel_oltr_resnet_MOZ_S2_111120.yaml

Stage 2: energy fine-tuning

python main.py --config ./configs/Stage_2/pslabel_oltr_energy_resnet_MOZ_S2_111620.yaml

Stage 2: deploying

python main.py --config ./configs/Stage_2/pslabel_oltr_energy_resnet_MOZ_S2_111620.yaml --deploy

Demo

A demo of this code can be found in [here] in CodeOcean.

License and Citation

The use of this software is released under BSD-3.

@article{10.1038/s42256-021-00393-0, 
year = {2021}, 
title = {{Iterative human and automated identification of wildlife images}}, 
author = {Miao, Zhongqi and Liu, Ziwei and Gaynor, Kaitlyn M and Palmer, Meredith S and Yu, Stella X and Getz, Wayne M}, 
journal = {Nature Machine Intelligence}, 
doi = {10.1038/s42256-021-00393-0}, 
pages = {885--895}, 
number = {10}, 
volume = {3}
}