ABSTRACT
In this paper, individual cattle instances are identified based on their characteristic features using hybrid of Mask R-CNN and abridged Residual Network (ResNet). In recent years, a hybrid of deep learning models is widely applied to image object identification and classification. To detect, segment and classify individual cattle instances for identification using Mask R-CNN algorithms, Keteku and Muturu cattle breeds were employed and concisely discussed in this paper. The datasets comprising a total of 1000 images belongs to the two cattle breeds: 600 Keteku images and 400 Muturu images. The content of the datasets are in image format, they were captured from the Northern part of Nigeria using a high resolution camera and smart phones. The images were augmented after preprocessing them to make up for the affected images caused by external factors such as variation in illumination, then, they were categorized. The work presented in this paper could be beneficial to researchers and farmers who make use of deep learning as essential tool for machine learning and artificial intelligence based models to monitor cattle’s production, weight estimation, body condition score and health related issues for sustainable livestock management.
Support the magazine and subscribe to the content
This is premium stuff. Subscribe to read the entire article.
