Skip to content

Latest commit

 

History

History
49 lines (29 loc) · 3.31 KB

README.md

File metadata and controls

49 lines (29 loc) · 3.31 KB

Satellite Image-to-Map Generation Project


Generates maps from satellite images.

The generated maps classify regions of the satellite image as Buildings, Roads, Water and Other.

Each region is color-coded a different shade of grey to differentiate them in the map.


data_collection/ Folder to extract data from Google Maps API. We can filter the map image generation by State in the getmapimages.py

model_generation/ Scripts used to generate training data and using training model. We used Weka to generate decforests2.model which is a Random Forest implementation. To generate prediction images, load map_data/ with satellite data of 800x800 satellite images(containing "satellite" in the filename), and then run python train_test.py. The map generation images from the model will be generated in the map_data/ folder. To change the model being used we have to use weka to generate a model and change the model location string in predict.java and recompile the class. Then re-run python train_test.py.

mapgen_results/ Results of test map generation runs from initial phase of model testing and evaluation. There are 3 runs displayed here.

**

Paper on Research Project

Automated Map Generation Project Paper

Sample Generated Results

enter image description here

enter image description here

enter image description here

enter image description here

enter image description here

enter image description here

enter image description here

enter image description here

enter image description here

enter image description here