Year: 2024
Type: article
Abstract: Conventional image-based techniques for discerning the sex of chicks have inherent drawbacks, such as the subjectivity involved in image selection and limited applicability to industrial contexts. In order to tackle these challenges, we employ videos in this study as an alternative to images, and present a more pragmatic approach that is suited to industrial applications. By leveraging an optimized PicoDet model, this methodology identifies telltale reflective attributes within the cloacae region of chicks. This approach also suggests that the sex of the chicks can be determined by calculating the proportion of male chick identifications in the video relative to the total number of images. Experimental findings demonstrate the superior performance of the proposed approach over the YOLO algorithm in terms of both cloacae and chick sex recognition. Optimal recognition efficiency is achieved when the aforementioned proportion falls within the range 60–70%. The accuracy rates for identifying female and male chicks were recorded as 90.34%, 91.33%, and 90.83%, respectively. The scheme developed in this study also achieves a reduction of 5.01% in model parameters, while the running time is shortened to less than 1 s, while maintaining comparable recognition efficiency to that of the PicoDet model. In summary, the method proposed in this paper exhibits enhanced proficiency in regard to recognizing both chick cloacae and their respective sexes. It successfully overcomes the limitations encountered by traditional image-based methodologies, and minimizes model space requirements. Furthermore, by harnessing the power of video, this approach has increased recognition accuracy and operational efficiency, ultimately improving the practicality and dissemination potential of this cutting-edge technology.
Source: Semina Ciências Agrárias
Authors: Keqiang Li, Yuqing Wang, Jiannan Yu, Xianglong Li
Institution: Hebei Normal University of Science and Technology
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: gold
Estimated APC paid: $133
OpenAlex ID: W4406119220