1. Deep Convolutional Neural Network for Reproductive Performance Selection in Rooster Breeding
- Author
-
Xuhong Lin, AiQiao Liu, Yifei Chen, and Yiping Sun
- Subjects
biology ,Computer science ,business.industry ,Rooster ,Phenotypic trait ,biology.organism_classification ,Machine learning ,computer.software_genre ,Convolutional neural network ,Task (project management) ,Breeder (cellular automaton) ,Artificial intelligence ,business ,computer ,Selection (genetic algorithm) - Abstract
There are three main disadvantages including time-consuming task, high cost and complex detection procedures in the semen quality measurement to heighten the roosters’ reproductive capability in breeder flocks. Another solution is to select the breeder roosters with fine phenotypic characteristics by humans, while it is also a considerably labor-intensive task and even increases the risk of zoonoses at a poultry farm. To solve these problems, this paper proposes a strategy that effective promoting factors applied to Progressive Multi-Granularity (PMG) network ensures the accuracy of entire image and improves the accuracy of fine-grained image. This strategy allows the basic networks boost the classification performance in the case of specific combination. Given the same images inputted into our model, two groups of questionnaires for practitioners and non-practitioners judging the fertility by the rooster’s phenotypic traits, the experimental results show that our method has raised the accuracy by almost 10% by comparison with the results of questionnaire survey.
- Published
- 2021
- Full Text
- View/download PDF