Skip to content
Image Classification on Satellite Imagery in Colombia via Keras and Tensorflow in Python
Python R
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data
logs
output
tools
.gitignore
README.md
gen_patches.py
predict.py
train_unet.py
unet_model.py

README.md

Keras and Tensorflow for creating a classified map

We are developing a pipeline using Python, Keras, and Tensorflow to classify satellite images from Planet. This work was performed with the Remote Sensing and Sustainability Lab. This served as the Capstone Project of my Professional Science Master's in GIS at Temple University.

The full report of this work is found in two parts on my portfolio. Thanks to Temple University's Remote Sensing and Sustainability Lab and reachsumit, who provided a fantastic unet example designed for performing image segmentation on satellite imagery.
Part 1
Part 2

  • train_unet.py is used to build the model. There are a number of configurable parameters such as number of Bands, number of classes, image size, patch size, and number of epochs.

  • predict.py is used to create predictions from an existing model, in this case classifying satellite images. It has a few configurable parameters, such as Image Directory, Image ID to predict against, and there is a debug flag that can enable more output to assist with troubleshooting.

  • gen_patches.py and unet_model.py are called by the two scripts above. It is not necessary to call these directly.

  • the tools directory contains a number of smaller utility scripts used in this research

You can’t perform that action at this time.