Abstract: Muscle fibers represent the fundamental structural units of skeletal muscle, and their morphology is closely linked to functional performance, adaptability to mechanical stress, and quality traits in both biomedical and agricultural contexts. During resistance training, skeletal muscle experiences mechanical tension that induces micro-damage, followed by repair and remodeling processes that enlarge the cross-sectional area (CSA) of fibers. This adaptive phenomenon, often referred to as supercompensation, underlies muscle hypertrophy, strength gains, and enhanced tolerance. In livestock science, fiber number, CSA, and diameter are critical determinants of meat quality, influencing pH, tenderness, protein composition, and water-holding capacity. Smaller fibers with lower CSA are associated with improved tenderness, whereas enlarged fibers often reduce meat quality. In medical research, quantitative analysis of fiber morphology provides diagnostic markers for neuromuscular disorders, muscle atrophy, and therapeutic efficacy, with CSA reduction serving as a hallmark of cachexia and age-related degeneration. Histological evaluation of muscle fibers commonly employs Hematoxylin-Eosin (HE) staining, a cost-effective and reproducible method that enables visualization of fiber boundaries and morphology. However, manual or semi-automated quantification of fiber size and density is labor-intensive and prone to subjectivity, particularly in high-density tissue sections. Recent advances in computer vision and deep learning have introduced segmentation models such as Cellpose, StarDist, RetinaMask, and FeatureNet, which improve efficiency but remain limited in accuracy for complex muscle structures. The Segment Anything Model (SAM) has attracted attention for its zero-shot generalization, yet its computational demands hinder large-scale deployment. Lightweight derivatives such as Fast Segment Anything Model (FastSAM), Mobile Segment Anything Model (MobileSAM), and Edge Segment Anything Model (EdgeSAM) offer improved inference speed and flexibility. This study developed an automated pipeline for muscle fiber quantification based on FastSAM, which integrates a YOLOv8-segmentation backbone with optimized pre- and post-processing strategies. Pig longissimus dorsi samples across eight developmental stages were collected, fixed, embedded, sectioned, and stained with HE. Images were acquired at 10× magnification using a Zeiss upright microscope and processed through grayscale conversion, histogram equalization, and Laplacian sharpening to enhance fiber boundaries. Large images were systematically cropped into overlapping sub-regions to preserve fiber integrity, followed by batch inference on an NVIDIA L40 graphics processing unit. Post-processing included removal of noise-dominated regions, reconstruction of fiber boundaries, and calculation of CSA and Feret diameters (minimum, maximum, and averaged values). Results were standardized in COCO format to facilitate reproducibility and integration with other platforms. A web-based interface, containerized via Docker and supported by NGINX, was implemented to enable user-friendly image upload, batch processing, visualization, and data export. Comparative evaluation against SAM, MobileSAM, and EdgeSAM demonstrated superior performance of FastSAM, with higher Dice coefficient, precision, recall, and F1 score, while maintaining lower inference latency than SAM. Correlation analysis revealed strong agreement between automated and manual measurements (r > 0.99), validating the reliability of the system. Application to twelve distinct muscle groups highlighted significant heterogeneity in CSA and diameter distributions, reflecting functional specialization and fiber type composition. Developmental analysis across weight stages confirmed progressive enlargement of fibers, consistent with established growth physiology. In conclusion, the proposed FastSAM-based framework provides a robust, efficient, and accurate solution for automated quantification of muscle fiber morphology in HE - stained sections. It overcomes limitations of manual and semi-automated methods, supports high - throughput analysis, and offers standardized outputs for downstream research. While current implementation does not differentiate fiber types, integration of immunohistochemical markers and multimodal learning is expected to enhance specificity in future studies. This approach establishes a scalable platform for biomedical diagnostics, livestock breeding, and meat quality assessment, bridging computational innovation with applied muscle biology.