Skylearn: High School Feasibility Analysis System Based on Satellite Imagery

Objective

Research project integrating Machine Learning and Deep Learning to analyze school images, covering the entire pipeline from data scraping, image acquisition, and feature extraction to clustering, classification, and model deployment.

Project Details

• Designed an end-to-end automated workflow for school feasibility analysis, processing 14,000+ school data and satellite images, significantly enhancing process efficiency.
• Automated scraping of school data and satellite imagery, reducing manual work from several weeks → 2 hours.
• Extracted thousands of image features using HOG, LBP, and deep learning models (VGG16, ResNet50, MobileNetV2).
• Automatically identified > 150 anomalies in school image data, ensuring cleaner and more accurate datasets.
• Clustered schools into two categories (“Feasible” and “Not Feasible”) to support strategic decision-making.
• Served as a **mentor for the research project led by Algae Desma Fridasary**, providing strategic direction, methodological guidance, and technical support in machine learning, computer vision, and data modeling.

Comments (2)