The performance of the YOLOX models was evaluated in two scenarios: object detection and classification. Confusion matrices for weed detection using YOLOX-nano and YOLOX-s are shown in Figures 5a and 5d, respectively. A comparison of weed true labels against None-predicted labels highlights the enhanced capability of the YOLOX-s model in detecting weeds that YOLOX-nano fails to detect. Furthermore, YOLOX-s achieved 13% and 16% higher accuracy in correctly detecting Nutsedge and Pigweed, respectively, albeit with a 7% reduction in accuracy for Purslane detection. Both models demonstrated high accuracy in classifying weed species, with YOLOX-nano showing a slight advantage over YOLOX-s in classifying Purslane (Figures 5b and 5e).