Scholarlink

<p>The objective of this research is to optimize the eleventh iteration of You Only Look Once (YOLOv11) by developing size-specific modified versions of the architecture. These modifications involve pruning specific layers and reconfiguring the main architecture of YOLOv11. Each proposed version is tailored to detect objects of specific size ranges, from small to large. To ensure proper model selection based on dataset characteristics, we introduced an object classifier program. This program identifies the most suitable modified version for a given dataset. The proposed models were evaluated on various datasets and compared with the original YOLOv11, YOLOv10, and YOLOv8 models. The experimental results highlight significant improvements in computational resource efficiency, with the proposed models maintaining the accuracy of the original YOLOv11. In some cases, the modified versions even outperformed the original model in terms of detection performance. Furthermore, the proposed models demonstrated reduced model sizes—each of the six proposed models showed a notable reduction compared to the original. Additionally, the required GFLOPs were reduced from 6.3&#xa0;MB (YOLOv11), 5.7&#xa0;MB (YOLOv10) and 8.1&#xa0;MB (YOLOv8) to just 3.8&#xa0;MB for the large model. All proposed models also achieved faster inference times, significantly reducing the time required to detect objects in images. Models weights and the object size classifier can be found in this repository (<a href="https://github.com/AREEG94FAHAD/yolov11">https://github.com/AREEG94FAHAD/yolov11</a>).</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

YOLOv11 optimization for efficient resource utilization

  • 期刊论文

  • 发表日期：

    2025年6月27日

查看全文
  • Areeg Fahad Rasheed,
  • Mahdi Zarkoosh

摘要

The objective of this research is to optimize the eleventh iteration of You Only Look Once (YOLOv11) by developing size-specific modified versions of the architecture. These modifications involve pruning specific layers and reconfiguring the main architecture of YOLOv11. Each proposed version is tailored to detect objects of specific size ranges, from small to large. To ensure proper model selection based on dataset characteristics, we introduced an object classifier program. This program identifies the most suitable modified version for a given dataset. The proposed models were evaluated on various datasets and compared with the original YOLOv11, YOLOv10, and YOLOv8 models. The experimental results highlight significant improvements in computational resource efficiency, with the proposed models maintaining the accuracy of the original YOLOv11. In some cases, the modified versions even outperformed the original model in terms of detection performance. Furthermore, the proposed models demonstrated reduced model sizes—each of the six proposed models showed a notable reduction compared to the original. Additionally, the required GFLOPs were reduced from 6.3 MB (YOLOv11), 5.7 MB (YOLOv10) and 8.1 MB (YOLOv8) to just 3.8 MB for the large model. All proposed models also achieved faster inference times, significantly reducing the time required to detect objects in images. Models weights and the object size classifier can be found in this repository (https://github.com/AREEG94FAHAD/yolov11).