papers/zoo-mobile-inference/sections/99-conclusion.tex

\section{Conclusion}

\label{sec:conclusion}

We have presented ZooMobile, a framework for deploying species recognition on mobile devices in conservation field settings.
Taxonomic Knowledge Distillation compresses models by 94\% while preserving 95\% of accuracy.
Species-specific mixed-precision quantization achieves sub-50ms inference on mid-range smartphones.
Condition-Aware Inference adapts to degraded field conditions, recovering up to 5.3 percentage points of accuracy under low-light conditions.

A 6-month field deployment across 8 national parks with 142 rangers demonstrated 84.7\% field-verified accuracy, enabling 23,400 real-time species identifications and directly supporting 47 anti-poaching decisions and 12 human-wildlife conflict interventions.
ZooMobile closes the deployment gap between laboratory-grade species recognition and the realities of conservation fieldwork.

Future work will extend ZooMobile to acoustic identification, develop continual learning mechanisms for adapting to new species encountered in the field, and integrate with ZooEd for in-field conservation training.

% ============================================================
% References
% ============================================================
\begin{thebibliography}{30}

\bibitem[Ahumada et~al.(2020)]{ahumada2020wildlife}
Ahumada, J.~A., Fegraus, E., Birch, T., et~al. (2020).
\newblock Wildlife Insights: A platform to maximize the potential of camera trap and other passive sensor wildlife data for the planet.
\newblock {\em Environmental Conservation}, 47(1):1--6.

\bibitem[Beery et~al.(2018)]{beery2018recognition}
Beery, S., Van~Horn, G., and Perona, P. (2018).
\newblock Recognition in Terra Incognita.
\newblock {\em Proceedings of ECCV}, pages 456--473.

\bibitem[Berger-Wolf et~al.(2017)]{bergerwolf2017wildbook}
Berger-Wolf, T.~Y., Rubenstein, D.~I., and Stewart, C.~V. (2017).
\newblock Wildbook: Crowdsourcing, computer vision, and data science for conservation.
\newblock {\em arXiv preprint arXiv:1710.08880}.

\bibitem[Dosovitskiy et~al.(2021)]{dosovitskiy2021image}
Dosovitskiy, A., Beyer, L., and Kolesnikov, A. (2021).
\newblock An image is worth 16x16 words: Transformers for image recognition at scale.
\newblock {\em Proceedings of ICLR}.

\bibitem[Han et~al.(2015)]{han2015learning}
Han, S., Pool, J., Tran, J., and Dally, W.~J. (2015).
\newblock Learning both weights and connections for efficient neural networks.
\newblock {\em Advances in Neural Information Processing Systems}, 28.

\bibitem[Hinton et~al.(2015)]{hinton2015distilling}
Hinton, G., Vinyals, O., and Dean, J. (2015).
\newblock Distilling the knowledge in a neural network.
\newblock {\em arXiv preprint arXiv:1503.02531}.

\bibitem[Howard et~al.(2019)]{howard2019searching}
Howard, A., Sandler, M., and Chen, B. (2019).
\newblock Searching for MobileNetV3.
\newblock {\em Proceedings of ICCV}, pages 1314--1324.

\bibitem[Jacob et~al.(2018)]{jacob2018quantization}
Jacob, B., Kligys, S., and Chen, B. (2018).
\newblock Quantization and training of neural networks for efficient integer-arithmetic-only inference.
\newblock {\em Proceedings of CVPR}, pages 2704--2713.

\bibitem[Liu et~al.(2022)]{liu2022convnet}
Liu, Z., Mao, H., Wu, C.~Y., et~al. (2022).
\newblock A ConvNet for the 2020s.
\newblock {\em Proceedings of CVPR}, pages 11976--11986.

\bibitem[Picek et~al.(2022)]{picek2022overcoming}
Picek, L., Sulc, M., Matas, J., et~al. (2022).
\newblock Overcoming practical issues of deep active learning with species identification.
\newblock {\em Ecological Informatics}, 70:101714.

\bibitem[Schneider et~al.(2020)]{schneider2020three}
Schneider, S., Taylor, G.~W., and Kremer, S.~C. (2020).
\newblock Three critical factors affecting automated image species recognition.
\newblock {\em Ecology and Evolution}, 10(7):3503--3517.

\bibitem[Stevens et~al.(2024)]{stevens2024bioclip}
Stevens, S., Wu, J., and Thompson, M.~J. (2024).
\newblock BioCLIP: A vision foundation model for the tree of life.
\newblock {\em Proceedings of CVPR}, pages 19412--19424.

\bibitem[Tan and Le(2019)]{tan2019efficientnet}
Tan, M. and Le, Q.~V. (2019).
\newblock EfficientNet: Rethinking model scaling for convolutional neural networks.
\newblock {\em Proceedings of ICML}, pages 6105--6114.

\bibitem[Terry et~al.(2020)]{terry2020automated}
Terry, J.~C.~D., Roy, H.~E., and August, T.~A. (2020).
\newblock Thinking like a naturalist: Enhancing computer vision of fine-grained classification.
\newblock {\em Methods in Ecology and Evolution}, 11(11):1360--1374.

\bibitem[Touvron et~al.(2021)]{touvron2021training}
Touvron, H., Cord, M., Douze, M., et~al. (2021).
\newblock Training data-efficient image transformers and distillation through attention.
\newblock {\em Proceedings of ICML}, pages 10347--10357.

\bibitem[Tuia et~al.(2022)]{tuia2022perspectives}
Tuia, D., Kellenberger, B., Beery, S., et~al. (2022).
\newblock Perspectives in machine learning for wildlife conservation.
\newblock {\em Nature Communications}, 13:792.

\bibitem[Van~Horn et~al.(2018)]{vanhorn2018inaturalist}
Van~Horn, G., Mac~Aodha, O., Song, Y., et~al. (2018).
\newblock The iNaturalist species classification and detection dataset.
\newblock {\em Proceedings of CVPR}, pages 8769--8778.

\bibitem[Van~Horn et~al.(2021)]{vanhorn2021benchmarking}
Van~Horn, G., Cole, E., and Beery, S. (2021).
\newblock Benchmarking representation learning for natural world image collections.
\newblock {\em Proceedings of CVPR}, pages 12884--12893.

\bibitem[Wang et~al.(2022)]{wang2022continual}
Wang, Q., Fink, O., Van~Gool, L., and Dai, D. (2022).
\newblock Continual test-time domain adaptation.
\newblock {\em Proceedings of CVPR}, pages 7201--7211.

\bibitem[Wood et~al.(2011)]{wood2011ebird}
Wood, C., Sullivan, B., Iliff, M., et~al. (2011).
\newblock eBird: Engaging birders in science and conservation.
\newblock {\em PLOS Biology}, 9(12):e1001220.

\bibitem[Yu et~al.(2022)]{yu2022metaformer}
Yu, W., Luo, M., Zhou, P., et~al. (2022).
\newblock MetaFormer is actually what you need for vision.
\newblock {\em Proceedings of CVPR}, pages 10819--10829.

\bibitem[Zhou et~al.(2022)]{zhou2022domain}
Zhou, K., Liu, Z., Qiao, Y., et~al. (2022).
\newblock Domain generalization: A survey.
\newblock {\em IEEE Transactions on Pattern Analysis and Machine Intelligence}, 45(4):4396--4415.

\bibitem[Antje Worring, Zoo Labs Foundation Research(2025a)]{zoo2025dso}
Antje Worring, Zoo Labs Foundation Research (2025a).
\newblock Decentralized Semantic Optimization: Collaborative knowledge curation on distributed networks.
\newblock {\em Zoo Improvement Proposal ZIP-001}.

\bibitem[Antje Worring, Zoo Labs Foundation Research(2025b)]{zoo2025network}
Antje Worring, Zoo Labs Foundation Research (2025b).
\newblock Zoo Network Architecture: A decentralized infrastructure for open science.
\newblock {\em Zoo Technical Report ZTR-001}.

\bibitem[Antje Worring, Zoo Labs Foundation Research(2026)]{zoo2026datacommons}
Antje Worring, Zoo Labs Foundation Research (2026).
\newblock Zoo Data Commons: A decentralized repository for open biodiversity datasets.
\newblock {\em Zoo Improvement Proposal ZIP-009}.

\end{thebibliography}