文献类型： 会议论文

第一作者： Rong Qian

作者： Rong Qian ¹ ; Lele Zhou ¹ ; Yue Yu ² ; Lei Xu ³ ;

作者机构： 1.Agricultural Economy and Information Research Institute, Anhui Academy of Agricultural Sciences, Hefei, China

2.Hefei Fengle Seed Co., Ltd, Hefei, China

3.Animal Husbandry and Veterinary Research Institute, Anhui Academy of Agricultural Sciences, Hefei, China

关键词： Training;Accuracy;Annotations;Semantic segmentation;Digital images;Semantics;Manuals;Signal processing;Cows;Benchmark testing

会议名称： International Conference on Intelligent Computing and Signal Processing

主办单位：

页码： 594-597

摘要： Achieving precise beef image segmentation is crucial for cattle breeding advancements. However, most conventional methods, relying solely on digital image processing techniques, fall short in delivering accurate results. This paper proposes a novel approach to address this limitation. We introduce a self-prompting guided Semantic Anything Module (SAM), enabling us to achieve semantic segmentation of beef within a designated region in the image. Remarkably, our proposed method is highly parameter-efficient, achieving promising results with limited annotations. Additionally, to facilitate training and evaluation, we have constructed a dedicated beef benchmark dataset. Our method demonstrates superior performance compared to existing approaches through extensive experimental evaluations.

分类号： tp3