Improving AI performance in wildlife monitoring through species and environment-specific training: A case study on desert Bighorn sheep
Motion-activated cameras are widely used to monitor wildlife, offering a non-intrusive and cost-effective means to collect high volumes of data. Artificial intelligence (AI) models can expedite image processing, but automated species classifications can be too inaccurate to meet end-users' need...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-11-01
|
| Series: | Ecological Informatics |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1574954125001888 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849694925524828160 |
|---|---|
| author | Owen S. Okuley Christina M. Aiello Will Glad Kyle Perkins Richard Ianniello Neal Darby Clinton W. Epps |
| author_facet | Owen S. Okuley Christina M. Aiello Will Glad Kyle Perkins Richard Ianniello Neal Darby Clinton W. Epps |
| author_sort | Owen S. Okuley |
| collection | DOAJ |
| description | Motion-activated cameras are widely used to monitor wildlife, offering a non-intrusive and cost-effective means to collect high volumes of data. Artificial intelligence (AI) models can expedite image processing, but automated species classifications can be too inaccurate to meet end-users' needs. This study evaluates how selection of data for model training influences AI detection of a focal species (desert bighorn sheep; Ovis canadensis nelsoni) across similar and novel locations. We compared two AI models: a species-specialist (deep_sheep) and a species-generalist (CameraTrapDetectoR), identified sources of bias, and retrained the specialist model using two datasets targeted toward biases associated with classification failure. Testing on 95,547 images from 36 water sources (5 novel) in the Mojave and Sonoran Deserts revealed the specialist model outperformed the generalist by 21.44 % in accuracy and reduced false negatives by 45.18 %. The specialist model was retrained first on site-representative data, then on both site-representative and extreme image-condition data. Retraining iterations consecutively reduced the false negative rate (36.94 % → 6.23 % → 4.67 %) and improved reliability across sites at the cost of a reciprocal increase in false positive rate (2.87 % → 15.22 % → 23.97 %) and variation. The site-representative retraining had the highest overall accuracy. Accuracy at out-of-sample sites remained comparable to the full dataset, though minor performance declines were observed. We found that specifying an AI's training to single-species classification and conditions within specific environments produced robust classification accuracy at minimal data requirements. By narrowing objectives while ensuring adequate training data variety, we achieved 89.33 % accuracy with a small fraction of the training data required by similar performing models. |
| format | Article |
| id | doaj-art-92d6fdfbcc5c45d08445a2f2f8e87790 |
| institution | DOAJ |
| issn | 1574-9541 |
| language | English |
| publishDate | 2025-11-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Ecological Informatics |
| spelling | doaj-art-92d6fdfbcc5c45d08445a2f2f8e877902025-08-20T03:19:56ZengElsevierEcological Informatics1574-95412025-11-018910317910.1016/j.ecoinf.2025.103179Improving AI performance in wildlife monitoring through species and environment-specific training: A case study on desert Bighorn sheepOwen S. Okuley0Christina M. Aiello1Will Glad2Kyle Perkins3Richard Ianniello4Neal Darby5Clinton W. Epps6Department of Fisheries, Wildlife and Conservation Sciences, Oregon State University, 104 Nash Hall, Corvallis, OR 97331, United States of America; Corresponding author at: College of Science, University of Texas at El Paso, Texas, TX 79902, USA.Department of Fisheries, Wildlife and Conservation Sciences, Oregon State University, 104 Nash Hall, Corvallis, OR 97331, United States of AmericaWhiting School of Engineering, John Hopkins University, 3400 N Charles St, Baltimore, MD 21218, United States of AmericaDepartment of Fisheries, Wildlife and Conservation Sciences, Oregon State University, 104 Nash Hall, Corvallis, OR 97331, United States of AmericaCalifornia Department of Fish and Wildlife, 787 N Main St, Bishop, CA 93514, United States of AmericaNational Park Service, Mojave National Preserve, 2701 Barstow Rd., Barstow, CA 92311, United States of AmericaDepartment of Fisheries, Wildlife and Conservation Sciences, Oregon State University, 104 Nash Hall, Corvallis, OR 97331, United States of AmericaMotion-activated cameras are widely used to monitor wildlife, offering a non-intrusive and cost-effective means to collect high volumes of data. Artificial intelligence (AI) models can expedite image processing, but automated species classifications can be too inaccurate to meet end-users' needs. This study evaluates how selection of data for model training influences AI detection of a focal species (desert bighorn sheep; Ovis canadensis nelsoni) across similar and novel locations. We compared two AI models: a species-specialist (deep_sheep) and a species-generalist (CameraTrapDetectoR), identified sources of bias, and retrained the specialist model using two datasets targeted toward biases associated with classification failure. Testing on 95,547 images from 36 water sources (5 novel) in the Mojave and Sonoran Deserts revealed the specialist model outperformed the generalist by 21.44 % in accuracy and reduced false negatives by 45.18 %. The specialist model was retrained first on site-representative data, then on both site-representative and extreme image-condition data. Retraining iterations consecutively reduced the false negative rate (36.94 % → 6.23 % → 4.67 %) and improved reliability across sites at the cost of a reciprocal increase in false positive rate (2.87 % → 15.22 % → 23.97 %) and variation. The site-representative retraining had the highest overall accuracy. Accuracy at out-of-sample sites remained comparable to the full dataset, though minor performance declines were observed. We found that specifying an AI's training to single-species classification and conditions within specific environments produced robust classification accuracy at minimal data requirements. By narrowing objectives while ensuring adequate training data variety, we achieved 89.33 % accuracy with a small fraction of the training data required by similar performing models.http://www.sciencedirect.com/science/article/pii/S1574954125001888Artificial intelligenceMachine learningTrail cameraCamera trapWildlife image classificationSpecies detection |
| spellingShingle | Owen S. Okuley Christina M. Aiello Will Glad Kyle Perkins Richard Ianniello Neal Darby Clinton W. Epps Improving AI performance in wildlife monitoring through species and environment-specific training: A case study on desert Bighorn sheep Ecological Informatics Artificial intelligence Machine learning Trail camera Camera trap Wildlife image classification Species detection |
| title | Improving AI performance in wildlife monitoring through species and environment-specific training: A case study on desert Bighorn sheep |
| title_full | Improving AI performance in wildlife monitoring through species and environment-specific training: A case study on desert Bighorn sheep |
| title_fullStr | Improving AI performance in wildlife monitoring through species and environment-specific training: A case study on desert Bighorn sheep |
| title_full_unstemmed | Improving AI performance in wildlife monitoring through species and environment-specific training: A case study on desert Bighorn sheep |
| title_short | Improving AI performance in wildlife monitoring through species and environment-specific training: A case study on desert Bighorn sheep |
| title_sort | improving ai performance in wildlife monitoring through species and environment specific training a case study on desert bighorn sheep |
| topic | Artificial intelligence Machine learning Trail camera Camera trap Wildlife image classification Species detection |
| url | http://www.sciencedirect.com/science/article/pii/S1574954125001888 |
| work_keys_str_mv | AT owensokuley improvingaiperformanceinwildlifemonitoringthroughspeciesandenvironmentspecifictrainingacasestudyondesertbighornsheep AT christinamaiello improvingaiperformanceinwildlifemonitoringthroughspeciesandenvironmentspecifictrainingacasestudyondesertbighornsheep AT willglad improvingaiperformanceinwildlifemonitoringthroughspeciesandenvironmentspecifictrainingacasestudyondesertbighornsheep AT kyleperkins improvingaiperformanceinwildlifemonitoringthroughspeciesandenvironmentspecifictrainingacasestudyondesertbighornsheep AT richardianniello improvingaiperformanceinwildlifemonitoringthroughspeciesandenvironmentspecifictrainingacasestudyondesertbighornsheep AT nealdarby improvingaiperformanceinwildlifemonitoringthroughspeciesandenvironmentspecifictrainingacasestudyondesertbighornsheep AT clintonwepps improvingaiperformanceinwildlifemonitoringthroughspeciesandenvironmentspecifictrainingacasestudyondesertbighornsheep |