How anmolrawat990/sen2sr-spectral-enhancement works
SEN2SR is a notebook-based remote-sensing research workflow for enhancing Sentinel-2 multispectral imagery, chiefly from 10 m to 2.5 m while preserving radiometry and spectral indices. Users prepare or obtain scenes and benchmark datasets, run pretrained Lite or Full inference, fine-tune with compound spatial/spectral losses, and compare outputs; optional cascaded experiments explore finer resolutions. The graph includes documented capabilities and the sampled WorldStrat extraction utility. Most notebook internals were not sampled, so their links reflect README statements rather than inspected calls.
Main entry points: 01_data_preparation Scene preparation · Train_SEN2SR_Lite_Compound_Loss.ipynb Lite fine-tuning · Train_SEN2SR_Full_Compound_Loss.ipynb Full fine-tuning · Train_SEN2SR_Cascaded_1.5m_VHR.ipynb 1.5m cascaded training
Components 18
Data preparation
- Benchmark datasets
notebooks/01_data_preparation - WorldStrat extractor
scripts/extract_worldstrat.py - Scene preparation
notebooks/01_data_preparation
Enhancement experiments
- Lite inference
notebooks/02_baseline_inference - Full inference
notebooks/02_baseline_inference - Compound loss method
notebooks/03_methodology_docs - Cascaded refinement method
notebooks/03_methodology_docs - Lite fine-tuning
notebooks/04_finetuning_experiments/Train_SEN2SR_Lite_Compound_Loss.ipynb - Full fine-tuning
notebooks/06_full_model_experiments/Train_SEN2SR_Full_Compound_Loss.ipynb - 1.5m cascaded training
notebooks/07_cascaded_1.5m_vhr_experiments/Train_SEN2SR_Cascaded_1.5m_VHR.ipynb - 1m refinement training
notebooks/05_ultra_highres_1m_experiments/Train_1Meter_Cascaded_Refiner.ipynb
Evaluation and results
- Lite comparison
notebooks/04_finetuning_experiments/SEN2SR_Baseline_vs_Finetuned_Comparison.ipynb - Full comparison
notebooks/06_full_model_experiments/SEN2SR_Full_Baseline_vs_Finetuned_Comparison.ipynb - 1.5m comparison
notebooks/07_cascaded_1.5m_vhr_experiments/SEN2SR_1.5Meter_MultiStage_Comparison.ipynb - 1m comparison
notebooks/05_ultra_highres_1m_experiments/SEN2SR_1Meter_4Way_Comparison.ipynb - Benchmark results
results
Other components
- Researcher
- Copernicus scenes
Connections 22
- Researcher → Copernicus scenes: downloads scenes
README.md - Researcher → Scene preparation: prepares data
README.md - Copernicus scenes → Scene preparation: supplies imagery
README.md - Scene preparation → Benchmark datasets: prepares datasets
README.md - WorldStrat extractor → Benchmark datasets: extracts archive
scripts/extract_worldstrat.py - Researcher → Lite inference: runs inference
README.md - Researcher → Full inference: runs inference
README.md - Researcher → Lite fine-tuning: fine-tunes model
README.md - Compound loss method → Lite fine-tuning: guides training
README.md - Researcher → Full fine-tuning: fine-tunes model
README.md - Compound loss method → Full fine-tuning: guides training
README.md - Cascaded refinement method → 1.5m cascaded training: guides refinement
README.md - Researcher → 1.5m cascaded training: runs experiment
README.md - Researcher → 1m refinement training: runs experiment
README.md - Lite fine-tuning → Lite comparison: compares fine-tuned model
README.md - Lite inference → Lite comparison: provides baseline
README.md - Full fine-tuning → Full comparison: compares fine-tuned model
README.md - Full inference → Full comparison: provides baseline
README.md - 1.5m cascaded training → 1.5m comparison: evaluates stages
README.md - 1m refinement training → 1m comparison: evaluates refinement
README.md - Lite comparison → Benchmark results: records metrics
- Full comparison → Benchmark results: records benchmarks
README.md