SAR stands for Synthetic Aperture Radar ,these images are high defination images which are rich in data and provide better insight to get knowldege of environmental changes,urban planning but due to lack of colors these images are not fully analysable so these images need to be colored , therefore we are developing a Deep Learning model for SAR image colorisation.
1- Python 2- conditional GANs 3- U-net (skip connections) 4- GDAL 5- Keras 6- OpenCV 7- Tensorflow 8- Flask
Here we used SAR images which are grayscale due to its high definition property and rich in data,due to this there is a lot of speckle noise
These are the colred SAR images which are generated by our trained model which reduces the speckle noise, image distortion
The colored SAR images can provide you a better interpretation of high definition data which are required for better analysis and environment monitering.