🔍 Objective
The paper aims to develop a machine learning-based system to automate the testing and classification of handloom sarees, replacing subjective human evaluation with objective, fast, and scalable methods.
🧵 Background
Traditionally, handloom sarees are assessed manually by experts based on visual, tactile, and experiential judgment, which is time-consuming, inconsistent, and labor-intensive. This leads to quality variability and inefficiency in production and quality control.
🛠️ Methodology
-
Image Dataset Collection:
-
High-resolution images of sarees with different yarn qualities and weaving patterns were collected.
-
Focus was placed on capturing warp and weft structures clearly.
-
-
Preprocessing:
-
Images were preprocessed using grayscale conversion, contrast enhancement, and noise reduction.
-
-
Feature Extraction:
-
Texture features were extracted using Gray Level Co-occurrence Matrix (GLCM).
-
Statistical parameters like contrast, correlation, energy, and homogeneity were computed.
-
-
Machine Learning Models Used:
-
SVM (Support Vector Machine)
-
KNN (K-Nearest Neighbors)
-
Naive Bayes
-
These models were trained to classify sarees into categories such as “Good,” “Average,” and “Poor” quality.
-
-
Evaluation Metrics:
-
Accuracy, precision, recall, and F1-score were used to evaluate model performance.
-
SVM outperformed the others with the highest accuracy.
-
📊 Results
-
SVM achieved over 90% accuracy in correctly classifying handloom sarees.
-
The automated system was faster, more consistent, and more scalable than human assessment.
🧠 Conclusion
The study demonstrates that machine learning can be effectively used to evaluate the quality of handloom sarees. This system can significantly benefit the textile industry, especially artisan clusters, by providing a reliable, repeatable, and automated quality check mechanism.
No comments:
Post a Comment