YieldNet: A Near-Zero-Cost YOLOv8n Enhancement for UAV-Based Real-Time Green Tomato Detection to Support Pre-Harvest Yield Forecasting

Abstract

Abstract Accurate pre-harvest yield forecasting of greenhouse tomatoes is essential for reducing post-harvest losses, with reliable detection of immature green tomatoes being the core challenge. However, these fruits are small, heavily occluded, and chromatically highly similar to foliage, making real-time detection from low-altitude UAV imagery extremely difficult, while onboard edge processors impose stringent power and weight constraints. To address this, we propose YieldNet, an ultra-lightweight framework that introduces near-zero-overhead enhancements to vanilla YOLOv8n: the backbone is replaced with ShuffleNetV2 to strengthen small-object representation; Efficient Channel Attention (ECA) modules are embedded after the P3–P5 layers in the neck to suppress leaf-background interference; and PIoU v2 loss is adopted to refine bounding-box regression for densely overlapped fruits via size-adaptive and non-monotonic focusing mechanisms. The model is rigorously validated on both a self-collected real-world UAV dataset comprising 600 low-altitude green-tomato images and a public multi-ripeness benchmark. Compared with the YOLOv8n baseline, YieldNet achieves relative improvements in mAP@50-95, Recall, and F1-score by 18.9%, 6.1%, and 5.8%, respectively, on the large dataset, and enhances Recall, F1-score, and Precision by relative gains of 4.3%, 4.0%, and 3.8%, respectively, on the small dataset, while increasing parameters only from 3.0\,M to 3.3\,M and reducing FLOPs from 8.1\,G to 8.0\,G. This work provides an efficient, readily deployable solution for high-precision real-time detection of immature green tomatoes on UAV platforms, enabling reliable pre-harvest yield estimation.

Plant phenotyping relevance

UAV画像から未成熟トマト果実を検出し、収量予測に用いるYOLOベースの画像解析手法を開発・複数データセットで検証しており、植物器官の表現型取得が中心です。

Code and data availability

The paper uses a self-built GreenTomato-UAV dataset (600 images) and a public 'Tomato-Recog' dataset, but no public deposit URL, repository, or availability statement for the self-built dataset, annotations, code, or trained YieldNet checkpoints appears in the supplied blocks. The public dataset is described only as 'a

No evidence-backed public reproduction asset is currently recorded.