Accurate weed identification is fundamental for precise weed spraying, a crucial aspect of modern agriculture (Tang et al., 2020). Leveraging computer vision technology, agriculture tasks like crop detection, phenotype analysis, and weed control have become more efficient (Lu and Young, 2020). Both traditional machine vision and Deep Learning (DL)-based methods excel in weed detection. Traditional approaches entail extracting weed characteristics such as colour, texture, and shape (Wang et al., 2019). Weeds pose a significant threat to rice cultivation, competing for resources and serving as hosts for pests, ultimately leading to yield loss and quality deterioration. Current weed control methods, primarily pesticide spraying, result in resource wastage and environmental harm (Swaminathan et al., 2023). Hence, precise spraying techniques are essential for effective weed suppression while minimizing pesticide usage and environmental impact. <!----> <!----> <!----> <!---->