| Popis: |
Non-chemical weeds control is a crucial aspect of sustainable organic agriculture. This paper explored the state-of-the-art deep learning networks applicability for weed detection and proposed a novel, simple deep learning model. The proposed network, CovWNET, is the second smallest network among the implementations compared in this study. It has roughly 1/3 times more parameters; however, it achieves 1.8% higher accuracy than the smallest network, MobileNetV2. Correspondingly, CovWNET has approximately five times fewer parameters and 2 % less accuracy than the best achieving model of DenseNet201. |
