Abstract Accurate monitoring of insect vectors is critical for managing Syndrome ‘basses richesses’ (SBR), a disease affecting sugar beet crops in Europe. This study presents a deep learning (DL) approach for the automated identification of Cixiidae planthoppers, the primary SBR vectors. Several DL architectures, including convolutional neural networks (CNN) and vision transformers, were benchmarked, leading to the selection of Inception-v3. This architecture was then used to develop two complementary models: the first to distinguish Cixiidae from other insect groups, and the second to classify species within Cixiidae, namely Pentastiridius leporinus , Hyalesthes obsoletus , and Reptalus spp. Discrimination among these species is essential, as they differ in their efficiency and role in transmitting SBR. The models were trained and validated on over 40,000 high-resolution insect images collected from sticky traps deployed in sugar beet and grapevine fields across Germany and Serbia between 2022 and 2024. Species labels were based on morphological identification and verified via DNA barcoding, and the resulting dataset was made publicly available. Statistical and visual model evaluation confirmed high performance and biological relevance. The first model achieved 94% overall accuracy with a 99% recall for Cixiidae, while the second model exceeded 99% overall accuracy. Performance was further evaluated on unseen test traps from 2024, showing high agreement between manual and model counts (R 2 > 0.9). Analysis of image resolution identified 1,600 dpi as the optimal balance between performance and computational efficiency. This study demonstrates the potential of DL for reliable SBR-vector monitoring and scalable integration into precision pest management strategies.
Okole et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: