Synthetic Data Generation for Fresh Fruit Bunch Ripeness Classification

In the field of fresh fruit bunch ripeness classification, the availability of annotated real-world datasets is often limited, posing challenges for developing accurate and robust classification models. To address this limitation, synthetic data generation techniques have emerged as a promising solution, offering the potential to augment dataset sizes and improve model performance. This study investigates the application of synthetic data for enhancing FFB ripeness classification. By leveraging simulation-based approaches, a diverse set of synthetic FFB images was generated, replicating variations in lighting conditions, object appearances, and environmental factors. The synthetic dataset was carefully designed to match the quantity of real data, ensuring a balanced comparison. Through extensive experiments, the performance of classification models trained on synthetic data was evaluated and compared with models trained solely on real data. The results demonstrated the efficacy of synthetic data in improving the accuracy and generalisation capability of the models. Furthermore, performance evaluation metrics, including mean average precision (mAP), were employed to assess the models’ performance across different ripeness levels. The findings highlight the potential of synthetic data for fresh fruit bunch ripeness classification and emphasise the importance of leveraging simulation-based techniques for generating high-quality synthetic datasets. This research contributes to the advancement of classification models in scenarios with limited real-world data availability, paving the way for improved accuracy and reliability in FFB ripeness classification.