Wetbird-DETR: Detection of Waterbirds in Satellite Wetlands of the Middle and Lower Reaches of the Yangtze River, China

The wetlands of the middle and lower reaches of the Yangtze River provide important habitats for migratory waterbirds, but repeated image-based monitoring in small satellite wetlands is difficult because birds are often distant, partly hidden by vegetation, and photographed against complex backgrounds. Here, the use of automated object detection was evaluated as a support tool for waterbird monitoring in these habitats. A lightweight detector, WetBird-DETR, was developed for field images collected in satellite wetlands. The detector was evaluated with a regional dataset from six satellite wetland sites in Anhui and Jiangsu Provinces, China, collected from June 2024 to June 2025 and containing 10 common waterbird species under varied light and background conditions. Supplementary evaluations were also conducted using two public bird datasets. WetBird-DETR maintained competitive detection performance while reducing the number of parameters and computational cost relative to the baseline detector. The results indicate that automated detection can support preliminary image screening, initial localization of birds, and expert review in repeated surveys of satellite wetlands. Broader validation across regions, seasons, habitat types, waterbird communities, and imaging systems will be necessary before the model can be used to support wider ecological inference.