Tools
ToolsZampetti, Andrea, Santini, Luca, Ferreiro-Arias, Iago, Paltrinieri, Laura, Ortiz, Iván, Cedeño-Panchez, Brayan A., Baltzinger, Christophe, Beirne, Christopher, Bowler, Mark T., Forget, Pierre Michel, Guilbert, Eric, Kemp, Yvonne J. M., Peres, Carlos A., Scabin, Andressa B., Whitworth, Andrew and Benítez-López, Ana (2026) Introducing TropiCam-AI: A taxonomically flexible automated classifier of Neotropical arboreal mammals and birds from camera-trap data. Methods in Ecology and Evolution. ISSN 2041-210X
Full text not available from this repository. (Request a copy)Abstract
Rapid, accurate assessment of arboreal vertebrates in tropical forests remains a bottleneck for large-scale biodiversity monitoring, due to the challenges and effort associated with traditional survey methods. To bridge this gap, arboreal camera-trapping is emerging as a promising way to observe otherwise elusive species, opening new avenues to advance behavioural ecology, community ecology and conservation biology. Yet, unlike ground-based camera-trapping, which has greatly benefited from machine learning innovations for automated species classification, equivalent tools for arboreal wildlife are almost non-existent. Here, we introduce TropiCam-AI, the first algorithm for automated species classification of Neotropical arboreal mammals and birds. We trained a deep learning architecture to recognize 84 taxa (63 species, 13 genera, 5 families and 3 orders), using a diverse set of camera-trap databases from Brazil (77,657 images), Peru (48,670 images), Costa Rica (44,857 images) and French Guiana (18,221 images). We also included in the training phase citizen science images from the iNaturalist platform (53,960 images) to increase generalizability and taxonomic coverage. Finally, we implemented a post-hoc aggregation strategy to further refine uncertain predictions by classifying them at higher taxonomic levels. TropiCam-AI reaches state-of-the-art testing accuracy of 95.0%, with the majority of taxa (50 out of 84) achieving >90% precision and recall, and 36 exceeding 95%. When allowing our model to return uncertain predictions at higher taxonomic ranks, predictive accuracy increased significantly, with a net gain of +4.3% in accuracy. Thanks to the carefully tuned data augmentation from iNaturalist, our model covers all 24 genera of New World monkeys and reliably classifies a wide array of other arboreal vertebrates. TropiCam-AI is available to use on the no-code AddaxAI platform, enabling a smooth and rapid uptake by a broad range of professionals working in the Neotropics. By reducing reliance on labour-intensive manual review of camera-trap data, TropiCam-AI empowers ecologists to accelerate data processing and ultimately enable effective assessment of Neotropical arboreal wildlife. This versatility can support not only targeted conservation actions but also fundamental ecological research in one of the world's most biodiverse yet under-surveyed ecosystems.
| Item Type: | Article |
|---|---|
| Additional Information: | Data Availability Statement: The TropiCam-AI algorithm is available and ready to use on the no-code AddaxAI platform (https://addaxdatascience.com/addaxai/). Further information on model usage can be found on the GitHub repository at https://github.com/andrewzamp/TropiCam-AI. All the codes used for model training and fine-tuning, along with the fully trained algorithm, iNaturalist images metadata and the full camera-trap dataset are also provided in a Zenodo repository at https://doi.org/10.5281/zenodo.17589042 (Zampetti et al., 2025). |
| Uncontrolled Keywords: | arboreal,artificial intelligence,camera-traps,canopy,machine learning,neotropics,primates,species classification,ecology, evolution, behavior and systematics,ecological modelling ,/dk/atira/pure/subjectarea/asjc/1100/1105 |
| Faculty \ School: | Faculty of Science > School of Environmental Sciences University of East Anglia Research Groups/Centres > Theme - ClimateUEA |
| UEA Research Groups: | Faculty of Science > Research Groups > Environmental Biology Faculty of Science > Research Centres > Centre for Ecology, Evolution and Conservation |
| Related URLs: | |
| Depositing User: | LivePure Connector |
| Date Deposited: | 01 Apr 2026 15:30 |
| Last Modified: | 05 Apr 2026 05:30 |
| URI: | https://ueaeprints.uea.ac.uk/id/eprint/102705 |
| DOI: | 10.1111/2041-210x.70213 |
Actions (login required)
 |
View Item |