Automated Cloth Quality Grading for E-Commerce Platform Using Machine Learning

In textile industry automation is very necessary for the better productivity and quality control of cloth inspection, especially in large manufacturing operations. The automated cloth quality detection and grading model that we propose in this research has low, medium, and high levels for cloth quality. It uses Python's ML libraries, and the classifiers used are random forest and logistic regression classifiers, along with the Adam optimizer for faster training. The data is labelled fabric image, pre-calibrated with label size and quality. Accuracy, precision, recall and F1 score are all measures by which we assess how good the model is at classification. The results have good accuracy and recall, which shows that the model does well at separating out different quality levels of fabric. The F1 score also confirms that the model is balanced in precision and recall, making it a good choice for field-based quality control applications. The automated system is sure to save you a lot of money and deliver better consistency and scalability, making it a great alternative to the manual inspections. In future updates, deep learning might be implemented to further optimize and upscale accuracy and scalability for larger textile datasets. Features such as high texture accuracy, medium colour detail, and thread densities of 52–299 are also used to distinguish between different quality states. There are tally and evaluate defects, from 0 to 9. Other key attributes are stretchability (low, medium, high), fabric thickness (thin to thick) and qualitative softness index (soft to rough). Moisture absorbency, pilling resistance and tear strength scores help narrow the classification; pilling resistance scores range from 9 to 10 and tear strength is between 5.0 and 49.4. This is evaluated by accuracy, precision, recall, and F1 score of the model. They show excellent classification with 92%accuracy, 90% precision, 88% recall, and 89% F1 score — the model is a well-balanced model for applications in the real world of quality control. The system is able to save money, ensure consistency, and expand as compared to manual checks. Future projects include deep learning to improve accuracy and scalability across textile datasets.