Deep learning approach for automatic segmentation of dirt on cattle skin using image data
Document Type
Conference Proceeding
Publication Title
2023 38th International Conference on Image and Vision Computing New Zealand (IVCNZ)
Publisher
IEEE
School
School of Science / Centre for Artificial Intelligence and Machine Learning (CAIML)
RAS ID
62730
Funders
Edith Cowan University
Abstract
Free-range cattle live and graze in soil and muddy pasture. This natural habitat easily makes the skin of cattle dirty and can lead to several issues not only for cattle but also for their raisers/farmers, who also have to meet certain requirements of hygienic meat and dairy products for consumers. Some of these issues include: (1) the stress experienced by cattle during the manual cleaning process by raisers, (2) the high labour cost of cleanliness and (3) the high amount of time required to clean a large number of cattle in a farm. These issues raise a need to develop an automatic and reliable cleanliness system to detect and segment dirt on the skin of cattle. In this paper, we propose a deep learning-based technique, called Deep Segmentation Network, to accomplish this challenging task. Our proposed deep learning-based system relies on image data to detect and segment dirt on the skin of cattle. We also propose baseline methods, which use ResNet50 and ResNet101 as the backbone models. We performed extensive experiments to validate the proposed approaches for the segmentation of dirt on cattle skin. Our experimental results demonstrate the superior performance of the proposed technique.
DOI
10.1109/IVCNZ61134.2023.10344224
Access Rights
subscription content
Comments
Islam, S. M. S., Shah, S. A. S., & Nguyen, C. D. M. (2023, November). Deep learning approach for automatic segmentation of dirt on cattle skin using image data [Paper presentation]. 2023 38th International Conference on Image and Vision Computing New Zealand (IVCNZ), Palmerston North, New Zealand. https://doi.org/10.1109/IVCNZ61134.2023.10344224