Dockerfile to create a docker image containing open source mritools for data handling, reconstruction and post-processing

This repository implements a neural network that upscales low-resolution brightness temperature data from AMSR2 satellite observations by 2×, 4×, or 8×. The model uses spatial attention mechanisms and residual learning to reconstruct high-resolution thermal fields while preserving physical consistency. Achieves 40-41 dB PSNR on 2× upsampling.