RPD: Learning Efficient Crops and Weeds for Field Semantic Segmentation in Drone Images (opens in new tab)

Efficient crop-weed segmentation is crucial for the perception needs of agricultural drones. However, natural illumination variations significantly alter the visual appearance of plants and pose major challenges for accurate crop-weed segmentation in field environments, particularly for small weeds. Existing deep learning (DL)-based methods mainly emphasize high-level semantic representations and may struggle to capture fine-grained structural cues essential for distinguishing weeds from crop...