Insects constitute the majority of animal species on Earth, yet recent studies indicate that their populations are experiencing a significant decline. Although these reports span multiple insect taxa and geographic regions, the evidence available to evaluate the scale of this decline remains limited. Monitoring insect populations is inherently difficult, as conventional methods are often labor-intensive, time-consuming, and inefficient. Recent advances in computer vision and deep learning offer promising avenues to address this global challenge. Cameras and other sensor-based technologies can provide continuous, noninvasive observations of insects across diurnal and seasonal cycles. In addition, automated laboratory imaging enables the capture of physical characteristics of specimens. Deep learning models trained on such data can be employed to estimate insect abundance, biomass, and species diversity. These models also allow the quantification of variation in phenotypic traits, behaviors, and ecological interactions. This paper explores the integration of deep learning and computer vision into entomological research, with a focus on improving cost-effective monitoring of insects and other invertebrates. We present examples of sensor-based monitoring systems and demonstrate how deep learning can be applied to large-scale datasets to extract ecological insights. Finally, we highlight four key areas critical to advancing this transformation: (1) validation of image-based taxonomic identification, (2) generation of sufficient training datasets, (3) development of publicly available, curated reference databases, and (4) integration of deep learning with molecular tools.
Sunitha et al. (Sat,) studied this question.