A critical challenge in fruit detection is the lack of annotated training data, which hampers the deployment of deep learning-based models in real-world conditions. To address this challenge, this study proposed a training data augmentation strategy based on Pix2Pix for improving the detection performance of YOLO v8s. In detail, we train a Pix2Pix model using the dataset to be tested, where the inputs are the converted three-channel grayscale images and the labels are the original RGB images. After training, we also converted the training images into three-channel grayscale inputs and then applied the trained Pix2Pix generator to perform style translation. Finally, we use the translated images to perform training data augmentation for YOLOv8s detection model. Our experimental results show that the proposed approach improves fruit detection performance across all metrics, with an average gain of 2.53% in precision, 2.08% in recall, 2.26% in F1-score, and 2.92% in mAP 50. These results demonstrate that the proposed data augmentation strategy effectively improves fruit detection performance under small-sample data conditions.