Add end-to-end example for reproducing TIRAuxCloud evaluation

[B-Mohid]

Resolves !
#9
This PR adds a minimal, end-to-end example to make it easier for new users to reproduce evaluation metrics on TIRAuxCloud pretrained models.
github

What this PR adds

examples/run_example_evaluation.py

Reads a small example config file (examples/example_saved_model_run.json).

Checks that:

dataset_dir and dataset_folder exist.

A pretrained model path is reachable (as in the paper models shared on Hugging Face).
github

Invokes the existing model_test.py entry point with the selected subset (landsat, landsatMA, or viirs).

Saves a metrics.json file with the main metrics reported by model_test.py (e.g. accuracy, precision, recall, F1, IoU) and optionally a confusion matrix.

examples/example_saved_model_run.json

A minimal configuration with placeholders and comments for:

dataset_folder (CSV lists for train/val/test).

dataset_dir (image samples directory).

device, cpuworkers, batch_size.

Model identifiers that correspond to one pretrained model from the Hugging Face repo.
github

examples/README.md

Step-by-step instructions:

Download the dataset and model weights from the TIRAuxCloud Hugging Face dataset page.

Populate the paths in example_saved_model_run.json.

Run:

python examples/run_example_evaluation.py --subset landsat

Inspect examples/output/metrics.json and (optionally) a saved confusion matrix image.

Main README.md updates

A short subsection under “Reproducing Model Metrics” that links to the examples/ folder for users who prefer a ready‑to‑run script.
github
Motivation

Lowers the barrier for users who are new to remote sensing / thermal cloud detection.

Provides a reusable evaluation component that can be plugged into broader AI frameworks for thermal satellite payload data analysis (e.g. for GSoC projects involving uncertainty, explainability, etc.).
github
+1
Notes / Questions

I assumed that calling model_test.py as a module (or via subprocess) is acceptable; if you prefer, I can refactor a small helper function in model_test.py to be imported directly.

If you’d like, I can extend the example to also save a few qualitative predictions (input TIR, mask, prediction) as PNGs under examples/output/ for quick visual inspection.
Happy to adjust the structure or naming to better fit the lab’s guidelines.