Browsing by Subject "camera traps"

This thesis investigates the application of machine learning (ML) technologies in the analysis of wildlife images captured by camera traps, focusing on its significance for ecology and conservation. With the advent of digital imaging and the increasing use of camera traps in wildlife monitoring, the volume of data thus generated has presented a significant challenge in terms of processing and analysis. This study aims to address this challenge by systematically reviewing the current state of ML applications in this field, identifying key technologies employed, and evaluating their effectiveness in various ecological and conservation tasks. Through an extensive literature review and analysis, the research reveals a strong preference for the use of convolutional neural networks (CNNs), particularly residual neural networks (ResNet), due to their ability to efficiently process and analyze large visual datasets. The study also highlights the primary ML tasks within this context, namely animal detection and species classification. Moreover, it identifies the main ecological objectives pursued through ML-assisted camera trap image analysis, emphasizing its frequent use in ecological assessments and wildlife population monitoring. The thesis identifies areas for improvement, such as addressing the underrepresentation of numerous taxa and enhancing the quality of the environmental assessments, and suggests directions for future research, including the development of more robust ML models capable of handling more diverse environments and datasets. By showcasing the potential of ML to revolutionize ecological research and conservation efforts, this study hopes to contribute to the understanding of how technology can be harnessed to preserve biodiversity and ensure more effective and efficient management of natural resources.