Transfer Learning-Driven Visual Garment Classification with Enhanced Feature Adaptation on Stitching Net

Abstract

In the modern era of fashion automation and textile manufacturing, the classification of garment types and stitching styles has become increasingly critical for tasks like quality assurance, inventory management, and automated product sorting. With the rising demand for intelligent systems in the apparel industry, there is a growing need for robust image-based garment analysis. Traditional approaches that depend heavily on human inspection or basic image scanning methods are no longer sufficient due to their inability to process high volumes of data with consistent accuracy. Manual classification systems suffer from several limitations including low scalability, inconsistent results, and inability to detect fine-grained stitching patterns. These methods lack automated learning, fail in complex visual conditions, and are inefficient for real-time operations. Traditional systems do not involve machine learning or feature extraction from convolutional networks; instead, they rely solely on visual cues and expert judgment, which introduces a high margin of error. This has created a critical gap that demands a more reliable, accurate, and scalable solution for garment classification. This project proposes an intelligent classification system using transfer learning through EfficientNetB0 for feature extraction, coupled with machine learning models like Deep Neural Networks (DNN), Perceptron, and a Proposed Multi layer perceptron (MLP) for dual-level classification such as identifying both garment category and stitching subclass. The entire workflow is integrated into a user-friendly Python-based Tkinter GUI that allows users to upload datasets, process images, train models, and predict garment types in real-time. By combining deep learning with traditional GUI design, the proposed system significantly improves classification accuracy, minimizes manual effort, and provides a scalable solution suited for smart garment industries.