Helper scripts and metric functions used by training and evaluation.

• get_training_vae_data.py
  • Builds Stage-I VAE training data from source datasets.
  • Expected output: single-channel H5 used by train_vae.py.
• calibrate_vae_scaling_factor.py
  • Utility for calibrating VAE latent scaling factor.
• metrics.py
  • Remote-sensing quality metrics used in validation/inference: Q4, Q8, SAM, ERGAS, SCC.

1. Prepare H5 training data with get_training_vae_data.py.
2. Train Stage-I VAE (train_vae.py).
3. Train Stage-II diffusion adapter (train_diffusion.py).
4. Evaluate with inference.py (which imports metrics from this folder).

Metric implementations here are treated as project ground truth during validation/inference.
If you replace them, keep formulas and input ranges consistent to avoid incomparable results.