https://github.com/norskregnesentral/naturinngrep
https://github.com/norskregnesentral/naturinngrep
Last synced: 8 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/norskregnesentral/naturinngrep
- Owner: NorskRegnesentral
- License: mit
- Created: 2020-03-05T12:19:53.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2020-03-13T09:58:44.000Z (over 5 years ago)
- Last Synced: 2025-01-10T16:46:12.933Z (10 months ago)
- Language: Python
- Size: 3.28 MB
- Stars: 2
- Watchers: 9
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
## Naturinngrep
Python code for applying convolutional neural networks (CNN) to Sentinel-2 data using python and PyTorch.
The code provides a simple framework for (1) converting Sentinel-2 data to an efficient format for deep learning, (2) training CNNs, and (3) applying trained networks to new data. We provide a simple example, classifiying road pixels in Sentinel-2 data. The code can easily be modified for other solving other problems by adding your own training data in the prepare_data.py-code.
### Setup:
- Make sure GDAL is installed.
- Download code and setup python:
```console
git clone https://github.com/NorskRegnesentral/naturinngrep.git
cd naturinngrep
virtualenv -p python3 env
source env/bin/activate
pip3 install -r REQUIREMENTS.txt
```
- You need to manually download the SAFE-files listed in s2_files*.txt. Also, rename the paths in these files to point to the location where the SAFE-files are downloaded. If you have your own files, these should be listed similarly in a text file.
### Main files:
- **prepare_data.py** - Move data from SAFE-format to a local storage with an efficient np.memmap-format suitable for training CNNs. Both training-data and test-data should be prepared. The file contains the processing configuration. Change the config paths to match your system.
- **train.py** - Train the CNN. In the script, the training and validation are specified. The trainied network are saved in the directory saved_models.
- **predict.py** - Apply the trained network to new data specified by the tiles list, and store as results as GEOtiff in the results directory. The
### Credit:
- UNet implementation by Jackson Huang: https://github.com/jaxony/unet-pytorch
- SAR geocoding library by the Norwegian Computing Center
### Authors:
Anders U. Waldeland (anders@nr.no), Arnt-Børre Salberg (salberg@nr.no), Øivind Due Trier (trier@nr.no)
Norwegian Computing Center (https://www.nr.no/)