Batik - Classification
Batik - Classification
Batik - Classification
CLIENT
Final Thesis (Academic)
Services
Machine Learning Computer Vision Web Development
Team
Ravi Aprillian (Lead Researcher)
Date
June 16, 2025
This project addresses the complexity of identifying Keraton Batik motifs, which often share similar patterns and textures. The primary goal was to build an automated classification system with high precision using deep learning technology, enabling users to upload digital images directly via a functional web interface.
This project addresses the complexity of identifying Keraton Batik motifs, which often share similar patterns and textures. The primary goal was to build an automated classification system with high precision using deep learning technology, enabling users to upload digital images directly via a functional web interface.
Solution
Solution
engineered a Feature Fusion architecture that merges global and local features. The core solution involves concatenating 2048 feature vector values from ResNet-50 with 20 manual feature values derived from HSV (Color), LBP (Texture), YCbCr (Color), Haralick (Texture) and Hu Moments (Pattern) analysis. This hybrid approach leverages the pattern recognition power of deep learning alongside domain-specific statistical data, resulting in a highly descriptive feature vector integrated into a Python-based web platform.
engineered a Feature Fusion architecture that merges global and local features. The core solution involves concatenating 2048 feature vector values from ResNet-50 with 20 manual feature values derived from HSV (Color), LBP (Texture), YCbCr (Color), Haralick (Texture) and Hu Moments (Pattern) analysis. This hybrid approach leverages the pattern recognition power of deep learning alongside domain-specific statistical data, resulting in a highly descriptive feature vector integrated into a Python-based web platform.
