How hemanthreddykunduru/change-detection-in-satellite-images works
This repository is a batch GeoTIFF object-detection pipeline. It reads images from cropped, runs sliced model inference, saves annotated images, and processes each detection into a crop and geographic coordinates before storing its metadata in PostgreSQL. The source shows starter.py orchestrating these stages and detection.py implementing detection and geospatial/image operations. No README is provided, so the graph reflects the sampled source and tree only.
Main entry points: starter.py Batch runner · detection.py Object detector · database.py Detection table writer · cropped GeoTIFF inputs
Components 10
Batch workflow
- Batch runner
starter.py - GeoTIFF inputs
cropped
Detection and geospatial processing
- Object detector
detection.py - SAHI inference
- Geo coordinate conversion
detection.py
Artifacts and persistence
- Annotated GeoTIFFs
detected - Object crops
objects - Detection table writer
database.py - PostgreSQL
Other components
- Pipeline operator
Connections 12
- Pipeline operator → Batch runner: starts batch
- Batch runner → GeoTIFF inputs: reads images
- Batch runner → Object detector: requests predictions
- Object detector → SAHI inference: runs sliced inference
- Batch runner → Object detector: requests visualization
- Object detector → Annotated GeoTIFFs: writes annotated images
- Batch runner → Object detector: requests crops
- Object detector → Object crops: writes image crops
- Batch runner → Object detector: requests coordinate conversion
- Batch runner → Geo coordinate conversion: converts pixel bounds
- Batch runner → Detection table writer: creates tables and saves detections
- Detection table writer → PostgreSQL: writes records