Manual strawberry quality inspection is physically demanding and presents a critical dilemma: aggressive removal leads to false positives (discarding healthy fruit), while lenient inspection risks false negatives (shipping defective fruit). Achieving optimal balance is difficult through human inspection alone. Due to fatigue, subjectivity, and inconsistency, human inspectors cannot simultaneously minimize both false positives and false negatives. This limitation results in either unnecessary waste or compromised product quality. This research aimed to develop an intelligent quality inspection assistance system capable of simultaneously reducing both false positive and false negative errors in strawberry sorting. We built an AI-assisted vision system using YOLO (You Only Look Once) computer vision, deployed on a Raspberry Pi 5 platform with integrated AI acceleration. This setup balances accuracy, speed, and affordability. The initial model successfully reduced both error types but suffered from overfitting due to annotation limitations. In the second version, we redesigned the annotation strategy, which significantly improved detection accuracy and model robustness. The solution remains low-cost (~$300 USD), energy-efficient, and easily scalable — ideal for smart farming applications in both developed and emerging markets.