Satellite Land-Use Classification (College Project)
Multi-class land-cover classification system using satellite imagery and deep learning - trained on multi-spectral data.
Project Snapshot
Role
Developer
Organization
JECRC University
Timeline
2023–2024
Status
Completed
Deployment
Academic / Research
Tech Stack
Impact & Results
- Multi-class land-cover mapping on satellite imagery
A college project demonstrating a multi-class land-cover classification system using multi-spectral satellite imagery and deep learning. This project laid the technical foundation that later informed my geospatial work at ISRO NRSC.
The Problem
Categorizing land use and land cover (e.g., urban, forest, water, agriculture) from satellite imagery is a fundamental task in remote sensing. It requires models capable of interpreting multi-spectral bands accurately.
The Solution
A classification and segmentation pipeline trained on multi-spectral remote sensing data to identify and map specific land cover types.
Architecture & Implementation
Modeling: Trained deep learning classification and segmentation models on multi-spectral satellite imagery.
Evaluation: Validated model accuracy against labeled ground truth data.
Data Used: Multi-spectral satellite images and labeled ground truth datasets.
Tech Stack: Python, PyTorch, QGIS, OpenCV, Rasterio, NumPy.
Challenges & Solutions
Data Sourcing: Finding high-quality, accurately labeled multi-spectral datasets available for students.
Pipeline Building: Building a robust deep learning training and evaluation pipeline entirely from scratch.
Compute: Training complex models with limited cloud compute resources.
Key Takeaways
Strong foundational knowledge of PyTorch and deep learning workflows.
Introduction to remote sensing fundamentals and multi-spectral imaging.
End-to-end ML pipeline creation from data loading to evaluation.