Synthetic and Derived Training Images for Campus Waste Detection: A Multi-Seed Evaluation with YOLOv8n

2026-07-21 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

A study evaluated the effectiveness of synthetic and derived images in enhancing a YOLOv8n detector for campus waste detection, specifically for bin-mounted camera views. Using a real dataset of 148 campus photographs, split into 86 for training, 31 for validation, and 31 for testing, researchers tested twelve joint-training configurations across seven principal settings with four matched seeds. The baseline real-only model achieved a mean mAP@0.5 of 0.691 [0.665, 0.722]. However, various synthetic image approaches, including background replacement (0.560 [0.499, 0.619]), isolated-object images (0.680 [0.644, 0.724]), and a full augmentation pool (0.487 [0.438, 0.537]), did not surpass this baseline. Hand-and-forearm composites also showed no reliable positive effect, with a corrected paired difference of +0.034 [-0.063, 0.199]. The findings indicate that none of the tested synthetic or derived image configurations improved upon the real-only model's performance, noting the 31-photo test set limits strong class-specific conclusions.

Key takeaway

For Machine Learning Engineers developing object detection models with limited real-world data, you should critically evaluate synthetic data generation strategies. This study suggests that simply adding synthetic or derived images, even with techniques like background replacement, may not improve YOLOv8n performance and can even degrade it. Prioritize rigorous multi-seed evaluation and consider the limitations of small test sets before integrating synthetic data into your production pipelines.

Key insights

Synthetic and derived images did not improve YOLOv8n waste detection performance over real-only data in this study.

Principles

• Data augmentation does not guarantee performance gains.
• Small test sets limit robust conclusions.
• Seed variation impacts model performance metrics.

Method

Evaluated YOLOv8n performance using 12 joint-training configurations, varying synthetic and derived image amounts and sources, repeated with four matched seeds for bootstrap percentile intervals.

In practice

• Evaluate synthetic data impact rigorously.
• Use multiple seeds for robust model comparisons.
• Ensure test sets are sufficiently large.

Topics

• YOLOv8n
• Object Detection
• Synthetic Data
• Data Augmentation
• Waste Detection
• Model Evaluation
• Computer Vision

Best for: AI Engineer, Research Scientist, AI Scientist, Machine Learning Engineer, Computer Vision Engineer

Related on AIssential

• Synthetic and Derived Training Images for Campus Waste Detection: A Multi-Seed Evaluation with YOLOv8n — Synthetic and derived training data did not improve YOLOv8n waste detection over real-only data, highlighting domain mismatch challenges.
• Making the Discrete Continuous: Synthetic RAW Augmentations for Fine-Grained Evaluation of Person Detection Performance in Low Light — Synthetic RAW augmentation can continuously evaluate low-light person detection, overcoming real data sparsity.
• ThorfinnYOLO: A hundred neural networks weekend — Dataset size and quality significantly outweigh hyperparameter tuning for object detection model performance.
• Training-Free Metrics for Synthetic Object Detection Data: A Proxy for Detector Performance — CCDM metrics provide a training-free, highly accurate proxy for evaluating synthetic object detection data utility, outperforming existing methods.
• Visual Prompting Meets Feature Reconstruction-Based Anomaly Detection with Dual-Teacher Supervision — Robust anomaly detection in varied real-world conditions is achieved by combining visual prompting, adaptive teacher models, and synthetic data.
• Balancing Real and Synthetic Data for CNN-based Masonry Crack Detection — Synthetic data can significantly reduce real data collection needs for CNN-based crack detection while maintaining or improving accuracy.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.