Cross-Generation Optimization of YOLOv26, YOLOv11, and YOLOv8 for Fine-Grained Small-Object Detection and Instance Segmentation in Complex Orchards
Cross-Generation Optimization of YOLOv26, YOLOv11, and YOLOv8 for Fine-Grained Small-Object Detection and Instance Segmentation in Complex Orchards
针对复杂果园细粒度小目标检测与实例分割的 YOLOv26、YOLOv11 与 YOLOv8 跨代优化研究
Abstract: Small-object detection and instance segmentation remain challenging in orchard environments because of green-on-green similarity, occlusion, and limited pixel representation of fine fruit anatomy. This study presents a cross-generation benchmark of Ultralytics YOLOv8, YOLOv11, and YOLOv26 for detecting and segmenting apple fruitlet, calyx, and peduncle structures for robotic orchard perception. 摘要： 在果园环境中，由于“绿中绿”的颜色相似性、遮挡问题以及细微果实结构像素表示不足，小目标检测与实例分割仍然极具挑战性。本研究对 Ultralytics YOLOv8、YOLOv11 和 YOLOv26 进行了跨代基准测试，旨在为果园机器人感知任务检测并分割苹果幼果、花萼和果梗结构。
Five model scales (n, s, m, l, and x) were evaluated under conventional 640 x 640 and small-object focused 960 x 960 training configurations, yielding 30 experiments. Increasing model capacity did not consistently improve accuracy. 研究在常规 640 x 640 和专注于小目标的 960 x 960 训练配置下，对五种模型规模（n、s、m、l 和 x）进行了评估，共进行了 30 组实验。结果表明，增加模型容量并不总是能带来精度的提升。
YOLOv11s-960 achieved the highest observed mask mAP@50:95 (0.402) and box mAP@50:95 (0.426), while YOLOv26s-960 achieved comparable values of 0.397 and 0.425 with only 10.37 M parameters and 34.1 GFLOPs. Peduncle remained the most challenging class. YOLOv11s-960 实现了观测到的最高掩码 mAP@50:95 (0.402) 和边界框 mAP@50:95 (0.426)；而 YOLOv26s-960 在仅有 10.37 M 参数和 34.1 GFLOPs 的情况下，达到了 0.397 和 0.425 的相当水平。其中，果梗（peduncle）依然是最具挑战性的类别。
Overall, compact-to-moderate YOLO models with small-object-focused training provided favorable accuracy efficiency trade-offs, establishing a practical benchmark for fine-grained agricultural robotics and orchard perception. 总体而言，采用小目标聚焦训练的紧凑型至中等规模 YOLO 模型在精度与效率之间取得了良好的平衡，为细粒度农业机器人和果园感知领域建立了一个实用的基准。
Authors: Ranjan Sapkota, Manoj Karkee 作者： Ranjan Sapkota, Manoj Karkee
Github Link: https://github.com/ (Note: Please refer to the original paper for the specific repository link) Github 链接： https://github.com/ （注：具体仓库链接请参考原论文）
Cite as: arXiv:2608.23636 [cs.CV] 引用格式： arXiv:2608.23636 [cs.CV]