Browsing by Author "Shin, Sook"

Now showing 1 - 4 of 4
  • Thumbnail Image
    Advancing Precision Agriculture Through AI and Statistical Modeling: Transforming Crop and Livestock Management
    Mann, Sahilpreet Singh (Virginia Tech, 2025-01-06)
    This thesis explores the application of Artificial Intelligence (AI), machine learning (ML), and statistical analysis to enhance agricultural practices, focusing on both livestock man- agement and plant biology. The first part investigates automated weight prediction of beef cattle using computer vision techniques, including YOLOv9 and InternImage with Cas- cade R-CNN for precise image segmentation. Advanced feature extraction methods utilizing ResNet, DenseNet, and ResNeXt are employed to develop ML and deep learning (DL) mod- els, providing a non-invasive alternative to traditional weight measurement techniques. The second part examines the regulation of the auxin response in Arabidopsis plants, focusing on epistatic interactions among auxin receptors. Through experimental assays and com- putational modeling, the study reveals synergistic effects that influence plant growth and development. The third part of the thesis characterizes the transcriptional specificity medi- ated by plant hormones using comprehensive data analysis, uncovering key insights into the gene regulation mechanisms influenced by auxin. Overall, the research integrates AI, ML, DL, and statistical methods to address critical challenges in agriculture and plant science, demonstrating improved predictive accuracy, enhanced understanding of hormonal signaling, and potential advancements in crop productivity and livestock management
  • Thumbnail Image
    AI-Driven Pig Monitoring System: Behavior and Weight Analysis
    Ranjan, Pranjal (Virginia Tech, 2024-12-12)
    This thesis advances automated pig monitoring through novel machine learning approaches in behavior analysis, weight prediction and forecasting. For behavior analysis, we introduce a preprocessing framework that addresses data leakage in time series analysis through non-class-based windowing and chronological sampling, achieving up to 15% improvement in accuracy over conventional methods. For current weight prediction, we develop an automated pipeline using the Segment Anything Model (SAM) with deep learning, where our Xception-Net architecture achieves a mean absolute percentage error of 7.42%. For weight forecasting, we propose multi-input deep learning architectures combining spatial and temporal features, achieving a mean absolute percentage error of 5.56%. These methods demonstrate robust performance in real-world conditions while minimizing animal stress and manual labor requirements, contributing significantly to precision livestock farming practices.
  • Thumbnail Image
    Machine Learning-Driven Optimization of Livestock Management: Classification of Cattle Behaviors for Enhanced Monitoring Efficiency
    Zhao, Zhuqing; Shehada, Halah; Ha, Dong; Dos Reis, Barbara; White, Robin; Shin, Sook (ACM, 2024-08-02)
    Monitoring cattle health in remote and expansive pastures poses significant challenges that necessitate automated, continuous, and real-time behavior monitoring. This paper investigates the effectiveness and reliability sensor-based cattle behavior classification for such monitoring, emphasizing the impact of intelligent feature selection in enhancing classification performance. To achieve this, we developed Wireless Sensor Nodes (WSN) affixed to individual cattle, enabling the capture of 3-axis acceleration data from five cows across varying seasons, spanning from summer to winter. Initially, we extracted a comprehensive set of 52 features, representing a broad spectrum of cow behaviors alongside statistical attributes. To enhance computational efficiency, we employed the Recursive Feature Elimination (RFE) method to distill 30 critical features by discarding redundant or less significant ones. Subsequently, these optimized features were utilized to train four machine learning (ML) models: Support Vector Machine (SVM), k-Nearest Neighbors (k- NN), Random Forest (RF), and Histogram-based Gradient Boosted Decision Trees (HGBDT). Notably, the HGBDT model demonstrated superior performance, achieving remarkable F1-scores of 99.01% for ’grazing’, 98.74% for ’ruminating’, 89.62% for ’lying’, 84.06% for ’standing’, and 91.87% for ’walking’. These findings underscore the potential of our approach to serve as a robust framework for precision livestock farming, offering valuable insights into enhancing cattle health monitoring in remote environments.
  • Thumbnail Image
    SegIt: Empowering Sensor Data Labeling with Enhanced Efficiency and Security
    Zhang, Zhen; Abraham, Samuel; Lee, Alex; Li, Yichen; Morota, Gota; Ha, Dong; Shin, Sook (ACM, 2024-08-02)
    SegIt is a novel, user-friendly, and highly efficient sensor data labeling tool designed to tackle critical challenges such as data privacy, synchronization accuracy, and memory efficiency inherent in existing labeling tools. While many current sensor data labeling tools provide free online services, they typically necessitate users to upload unlabeled sensor data, alongside video or audio references, to cloud storage for labeling. Nevertheless, such third-party storage exposes user data to potential security risks. SegIt, an innovative open-source tool, provides a software solution for tagging unlabeled sensor data directly on a local computer, ensuring enhanced accuracy, convenience, and, most importantly, data security.