SDQM: Synthetic Data Quality Metric for Object Detection Dataset Evaluation

The article presents the Synthetic Dataset Quality Metric (SDQM), designed to evaluate the quality of synthetic datasets for object detection tasks without requiring model training to converge. SDQM demonstrates a strong correlation with the mean average precision (mAP) scores of the YOLOv5 model, outperforming previous metrics, and provides actionable insights for dataset quality improvement. This metric addresses the challenge of efficiently generating and selecting synthetic datasets, which is crucial for practitioners working with resource-constrained object detection models.