SAR image denoising using deep learning

This post is about denoising of SAR image using sevaral deep learning methods.

MoNET_pytorch

Errors

1. RuntimeError: CUDA out of memory.

Tried to allocate 7.68 GiB (GPU 0; 47.37 GiB total capacity; 38.60 GiB already allocated; 7.07 GiB free; 38.62 GiB reserved in total by PyTorch)

SAR-CNN

Build environment in AutoDL

1. Create new conda environment: conda create --name sarcnn python=3.7

2. Activate sarcnn environment: source activate sarcnn

3. Optional: install mumba: conda install conda-forge/label/cf202003::mamba

4. Install numpy: conda install numpy=1.16.5

5. Install tensorflow-gpu: conda install tensorflow-gpu=1.13.1

6. Install pillow: conda install anaconda::pillow

Run SAR-CNN scripts

1. python main.py --real_sar=1 --test_dir='/autodl-fs/SAR-CNN-test/test_real/'

2. chmod -R 755 ~/autodl-fs/SAR-CNN-test/

3. (sarcnn) root@autodl-container-dc124d9a08-a11e31e5:~/autodl-fs/SAR-CNN-test# python main.py --real_sar=1 --test_dir='~/autodl-fs/SAR-CNN-test/test_real/'

4. Upload real Sentinel1-A SAR data into folder /root/autodl-fs/SAR-CNN-test/data/real

5. Run script /root/autodl-fs/SAR-CNN-test/main.py