KG-DBAT: Knowledge-guided dual-branch attention transformer with IoU-aware recalibration for environmental pollen detection

Key Points

• Enhanced accuracy in environmental pollen detection was achieved through the use of a dual-branch attention transformer.
• The model integrates IoU-aware recalibration, significantly improving detection metrics in real-time applications.
• Knowledge-guided methods were applied to refine the attention mechanisms, specifically targeting pollen classification and recognition.
• The findings suggest potential applications in ecological monitoring and public health, indicating a substantial impact on environmental studies.