Part 1: Development of an AI/ML model to generate high (~30 cm) resolution lunar terrain image from medium/low (5 m / 10 m) resolution terrain image, and evaluate its accuracy; the model will be amenable to further training and improvement.
Part 2: To generate a global lunar atlas (digital) with the help of the AI/ML model, based on the medium / low resolution data available. The AI/ML model to be developed by using the publically available overlapping data of OHRC (~30 cm resolution) and TMC-2 (5 m /10 m resolution) payload data on board Chandrayaan-2 Orbiter.
-
Final Atlas https://drive.google.com/drive/folders/1AlsNVxCAYomUHkd641Qv8PnI_gg1RQlV
-
Please find the high resolution
.tifstitched images produced after super resolution on TMC images using model-a: SRGAN and model-b: SWIN-IR in the google drive link here.
README
./pipeline.sh
part-i-ai-model/
- README
- model-a-srgan/ - based on gan modelling (srgan)
- README
- train and test scripts
- model-b-swinir/ - based on transformer arch (swinir)
- README
- train and test scripts
parti-ii-lunar-atlas-stitching/
- README
- assets/
- matlab code
dataset-cleaning-and-analysis/
- README
- python notebooks
- scripts
- assets/ (contain images of analysis and findings)
- SwinIR: Image Restoration Using Swin Transformer
@article{liang2021swinir,
title={SwinIR: Image Restoration Using Swin Transformer},
author={Liang, Jingyun and Cao, Jiezhang and Sun, Guolei and Zhang, Kai and Van Gool, Luc and Timofte, Radu},
journal={arXiv preprint arXiv:2108.10257},
year={2021}
}
- Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
@InProceedings{srgan,
author = {Christian Ledig, Lucas Theis, Ferenc Huszar, Jose Caballero, Andrew Cunningham, Alejandro Acosta, Andrew Aitken, Alykhan Tejani, Johannes Totz, Zehan Wang, Wenzhe Shi},
title = {Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network},
booktitle = {arXiv},
year = {2016}
}