In a significant academic endeavor, a team of esteemed faculty members from the Bangladesh University of Business and Technology (BUBT) presented their research project on September 20th , 2023. The research project, titled “Utilizing EfficientNet for Sheep Breed Identification in Low-Resolution Images,” marks a remarkable pace in the field of sheep farming
Funded by BUBT’s Business Research and Innovation Center (BRIC), the collaborative research team comprised Md. Mijanur Rahman, Assistant Professor and Principal Researcher from the Department of Computer Science and Engineering (CSE), Md. Masudul Islam, Assistant Professor and Co-researcher from the same department, and Mr. Galib Md. Sharhriar Himel, Co-researcher from another university.
The event, organized by the Institutional Quality Assurance Cell (IQAC) of BUBT, was graced by the presence of distinguished individuals including Prof. Md. Fayyaz Khan, Vice Chancellor of BUBT, Prof. Santi Narayan Ghosh, Zaved Mannan, Additional Director of IQAC, the Chairman of the CSE department, and esteemed faculty members.
The research project aimed to address a crucial challenge in the sheep farming industry – the accurate identification of sheep breeds. Recognizing sheep breeds is of immense importance to farmers, but it often requires specialized knowledge and can be a daunting task, especially for novices.
To tackle this challenge, the research team proposed an innovative solution: the adoption of a Convolutional Neural Network (CNN) model capable of swiftly and accurately identifying sheep breeds from low-resolution images. This model bridges the gap between expert-level breed identification and practical, on-the-go accessibility.
The team used a dataset of 1680 facial images, each representing one of four distinct sheep breeds, as the foundation for training and testing their CNN model. After rigorous experimentation, they found that EfficientNetB5, a cutting-edge deep learning model, outperformed others with an impressive accuracy rate of 97.62% on a 10% test split.
This outcome signifies the potential applicability of their approach in real-world scenarios. The classification model they developed not only assists sheep farmers in distinguishing between different breeds but also enables more precise assessments and sector-specific classification for various businesses within the industry.
Furthermore, the user-friendly nature of the model, deployable through handheld devices, democratizes advanced technology for farmers with limited expertise. It offers the potential for informed decision- making, targeted breeding programs, optimized resource allocation, and enhanced flock management.
In summary, this groundbreaking research underscores the significance of automating sheep breed recognition and introduces an efficient CNN-based model for the task. Through rigorous evaluation, the team has demonstrated the superior performance of their model, heralding accessible and expert-level breed identification.
This technology promises to empower farmers, foster industry advancements, and promote efficient agricultural practices in the sheep farming sector, contributing to its overall growth and sustainability.