w610001795 / CNN-SARNet

Python 2.7.12 : Use Deep Learning (Tensorflow) to classify SAR image, especially for SAR image of ocean

CNN-SARNet

Python 2.7.12
Tensorflow
The goal of this project is to provide a way to classify the small image of SAR in ocean, especially two-category.

Validation Result

accuracy Top 1 P = 0.9359

To Run the Project

main() is in the COSI.py, you can run it to have a train, and you should ensure that other files and COSI.py are in the same folder.
Command line statement：

python COSI.py --mode=train --max_steps=50006 --eval_steps=100 --save_steps=2000
python COSI.py --mode=validation
……

project structure

COSI.py ：includes the main functions and FLAGS
cnnnet.py ：includes the structure of the CNN-SARNet
DI.py ：data read, preprocessing and enhancement
log.py ：bulid the logger files of the project
train.py ：algorithm of training network
validation.py：validation method
inference.py ：inference for single test images

Inquiring

If you had some questions, you would email to me : [[e-mail]]
If you are also committed to applying Deep Learning to remote sensing (especially SAR ) research, you could also contact me for academic exchange.