With the advancement of computer vision technology, deep learning-based fruit detection techniques have been widely applied in modern orchards. However, these fruit detection models rely on fully supervised learning methods, which require large-scale annotated datasets to support model construction. Currently, many widely used open-source annotation tools heavily depend on manual labeling, making the process tedious and time-consuming. To address this issue, this study integrates previous research to develop FruitLabel, an efficient web-based orchard data annotation tool. By leveraging domain adaptation transformation and a pseudo-label self-learning algorithm, FruitLabel enables fully automated annotation of various fruit types without any manual intervention. This approach significantly improves annotation efficiency and reduces the high cost of dataset construction.