Optimizing the Accuracy and Efficiency of Camera Trap Image Analysis: Evaluating AI Model Performance and a Semi-Automated Workflow.
Saved in:
| Title: | Optimizing the Accuracy and Efficiency of Camera Trap Image Analysis: Evaluating AI Model Performance and a Semi-Automated Workflow. |
|---|---|
| Authors: | Hitchcock, Kelly1,2 (AUTHOR) kelly.hitchcock@ntu.ac.uk, Tollington, Simon1,2,3 (AUTHOR), Yarnell, Richard W.1,3 (AUTHOR), Williams, Leah J.2,4 (AUTHOR), Hamill, Kat1 (AUTHOR), Fergus, Paul2,4 (AUTHOR) |
| Source: | Remote Sensing. Feb2026, Vol. 18 Issue 3, p502. 20p. |
| Subjects: | Workflow, Biological classification, Wildlife photography, Citizen science, Wildlife monitoring, Biodiversity monitoring |
| Abstract: | Highlights: What are the main findings? Initial Conservation AI UK Mammals model outputs demonstrated high precision (>0.80) for foxes (Vulpes vulpes) and hedgehogs (Erinaceus europaeus) but low recall (<0.50) for hedgehogs. Following retraining, AI model performance improved substantially. However, discrepancies between AI and human classifications remained statistically significant, indicating that human accuracy still outperformed that of the AI model. Recall scores for hedgehogs also remained low despite these improvements. What are the implications of the main findings? We present a semi-automated, three-step workflow incorporating an AI generalist object detector, an AI species-specific classifier, and manual validation as an alternative image classification method that accelerates camera trap data analysis whilst maintaining classification accuracy. The findings provide baseline performance estimates of Conservation AI's UK Mammals model and highlight the importance of continuous AI model training, the value of citizen science in expanding training datasets, and the need for adaptable workflows in camera trap studies. The widespread adoption of camera trap surveys for wildlife monitoring has generated a substantial volume of ecological data, yet processing constraints persist due to the time-consuming process of manual image classification and the reliability of automated systems. This study assesses the performance of Conservation AI's UK Mammals model in classifying three species—Western European hedgehogs (Erinaceus europaeus), red foxes (Vulpes vulpes), and European badgers (Meles meles)—from a subsample of 234 records from camera trap images collected through a citizen science initiative across residential gardens. This analysis was repeated after retraining the model to assess improvement in model performance. Initial model outputs demonstrated high precision (>0.80) for foxes and hedgehogs but low recall (<0.50) for hedgehogs, with the lowest recall probability of 0.12 at the 95% confidence threshold (CT). Following retraining, model performance improved substantially across all metrics, with average F1-scores (weighted average of precision and recall across the three species tested) improving at all CTs, though discrepancies with human classifications remained statistically significant. Based on performance results from this study, we present a semi-automated, three-step workflow incorporating an artificially intelligent (AI) generalist object detector (MegaDetector), an AI species-specific classifier (Conservation AI), and manual validation. Where privacy concerns restrict citizen science contributions, our pipeline offers an alternative that accelerates camera trap data analysis whilst maintaining classification accuracy. The findings provide baseline performance estimates of Conservation AI's UK Mammals model and present an approach that offers a practical solution to improve the efficiency of using camera traps in ecological research and conservation planning. We also highlight the importance of continuous AI model training, the value of citizen science in expanding training datasets, and the need for adaptable workflows in camera trap studies. [ABSTRACT FROM AUTHOR] |
| Copyright of Remote Sensing is the property of MDPI and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Highlights: What are the main findings? Initial Conservation AI UK Mammals model outputs demonstrated high precision (>0.80) for foxes (Vulpes vulpes) and hedgehogs (Erinaceus europaeus) but low recall (<0.50) for hedgehogs. Following retraining, AI model performance improved substantially. However, discrepancies between AI and human classifications remained statistically significant, indicating that human accuracy still outperformed that of the AI model. Recall scores for hedgehogs also remained low despite these improvements. What are the implications of the main findings? We present a semi-automated, three-step workflow incorporating an AI generalist object detector, an AI species-specific classifier, and manual validation as an alternative image classification method that accelerates camera trap data analysis whilst maintaining classification accuracy. The findings provide baseline performance estimates of Conservation AI's UK Mammals model and highlight the importance of continuous AI model training, the value of citizen science in expanding training datasets, and the need for adaptable workflows in camera trap studies. The widespread adoption of camera trap surveys for wildlife monitoring has generated a substantial volume of ecological data, yet processing constraints persist due to the time-consuming process of manual image classification and the reliability of automated systems. This study assesses the performance of Conservation AI's UK Mammals model in classifying three species—Western European hedgehogs (Erinaceus europaeus), red foxes (Vulpes vulpes), and European badgers (Meles meles)—from a subsample of 234 records from camera trap images collected through a citizen science initiative across residential gardens. This analysis was repeated after retraining the model to assess improvement in model performance. Initial model outputs demonstrated high precision (>0.80) for foxes and hedgehogs but low recall (<0.50) for hedgehogs, with the lowest recall probability of 0.12 at the 95% confidence threshold (CT). Following retraining, model performance improved substantially across all metrics, with average F1-scores (weighted average of precision and recall across the three species tested) improving at all CTs, though discrepancies with human classifications remained statistically significant. Based on performance results from this study, we present a semi-automated, three-step workflow incorporating an artificially intelligent (AI) generalist object detector (MegaDetector), an AI species-specific classifier (Conservation AI), and manual validation. Where privacy concerns restrict citizen science contributions, our pipeline offers an alternative that accelerates camera trap data analysis whilst maintaining classification accuracy. The findings provide baseline performance estimates of Conservation AI's UK Mammals model and present an approach that offers a practical solution to improve the efficiency of using camera traps in ecological research and conservation planning. We also highlight the importance of continuous AI model training, the value of citizen science in expanding training datasets, and the need for adaptable workflows in camera trap studies. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 20724292 |
| DOI: | 10.3390/rs18030502 |
