Satellite-Image-Classification

Introduction

Train AI models on satellite image dataset to classify different types of land.

How to run

Run train.ipynb to train models on satellite image dataset.

Dataset

Segment satellite imagery into small blocks, and annotate type labels of blocks.

Demo dataset

We provide a small dataset in the repository. You can directly unzip it and see the following folder structure.

Satellite-Image-Classification/
├── dataset
|   ├── industry
|   |   ├── xxx.jpg
|   |   └── ...
|   ├── agriculture
|   └── residence

Create your own dataset

1. Get a satellite imagery in .tiff format.
2. Use Global Mapper to split the tiff file into small blocks in .jpg format.
3. Manually classify these images and create a folder structure like above.

Models

Models are listed in the folder models, which is imported from pytorch-cifar .

References

• pytorch-cifar