eval/plot.py

"""

eval/plot.py



Generate comparison figures for all cloud-removal methods.



The script reads from the cleaned-up directory layout created by migrate.py:



    visualization/

    ├── data/

    │   ├── Sen2_MTC_New/

    │   │   ├── GT/        {id}.png

    │   │   └── inputs/    {id}_A1.png  {id}_A2.png  {id}_A3.png

    │   └── Sen2_MTC_Old/

    │       ├── GT/

    │       └── inputs/

    └── results/

        ├── Sen2_MTC_New/{method}/{id}.png

        └── Sen2_MTC_Old/{method}/{id}.png



Usage

-----

# Generate the exact figures that appear in the paper:

    python plot.py --paper-samples



# Generate paper figures for one dataset only:

    python plot.py --paper-samples --dataset Sen2_MTC_New

    python plot.py --paper-samples --dataset Sen2_MTC_Old



# Generate a figure for any arbitrary sample ID:

    python plot.py --dataset Sen2_MTC_New --id T12TUR_R027_55



# List all available sample IDs for a dataset:

    python plot.py --dataset Sen2_MTC_New --list



# Custom output directory:

    python plot.py --paper-samples --out-dir /path/to/figures

"""

from __future__ import annotations

import argparse
import os
from glob import glob
from typing import Optional

import matplotlib
import matplotlib.pyplot as plt
import numpy as np

matplotlib.rcParams["font.family"] = "Times New Roman"

# ---------------------------------------------------------------------------
# Paths
# ---------------------------------------------------------------------------
# eval/ is one level below the project root
ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))

# ---------------------------------------------------------------------------
# Constants
# ---------------------------------------------------------------------------
DATASETS = ["Sen2_MTC_Old", "Sen2_MTC_New"]

# Display order in the 4×4 grid (row-major, after the 4 input/GT panels)
METHODS: list[str] = [
    "mcgan",
    "pix2pix",
    "ae",
    "stnet",
    "dsen2cr",
    "stgan",
    "ctgan",
    "crtsnet",
    "pmaa",
    "uncrtaints",
    "ddpmcr",
    "diffcr",
]

METHOD_LABELS: list[str] = [
    "MCGAN",
    "Pix2Pix",
    "AE",
    "STNet",
    "DSen2-CR",
    "STGAN",
    "CTGAN",
    "CR-TS-Net",
    "PMAA",
    "UnCRtainTS",
    "DDPM-CR",
    "DiffCR [Ours]",
]

INPUT_LABELS: list[str] = [
    r"Cloudy $T_1$",
    r"Cloudy $T_2$",
    r"Cloudy $T_3$",
    "Ground-Truth",
]

ALL_LABELS: list[str] = INPUT_LABELS + METHOD_LABELS

# Some methods in the Old dataset store outputs with a horizontal flip
# relative to the other methods' spatial convention.  We correct for display.
FLIP_H_FOR_DISPLAY: dict[str, set[str]] = {
    "Sen2_MTC_Old": {"diffcr"},
}

# The exact sample IDs used in the paper figures
PAPER_SAMPLES: dict[str, list[str]] = {
    "Sen2_MTC_New": ["T12TUR_R027_55"],
    "Sen2_MTC_Old": ["42WVD_70008000", "14SQB_20006000"],
}


# ---------------------------------------------------------------------------
# I/O helpers
# ---------------------------------------------------------------------------


def _find_input(inputs_dir: str, sample_id: str, channel: str) -> Optional[str]:
    """Locate {id}_A{1|2|3}.png in *inputs_dir*."""
    direct = os.path.join(inputs_dir, f"{sample_id}_{channel}.png")
    if os.path.exists(direct):
        return direct
    # Fallback – glob for any file containing the id and channel tag
    hits = glob(os.path.join(inputs_dir, f"{sample_id}*{channel}*"))
    return hits[0] if hits else None


def _load(path: str, flip_h: bool = False) -> np.ndarray:
    """Load an image as float [0,1] RGBA/RGB via matplotlib.



    matplotlib.imread returns:

    - PNG: float32 [0,1]  (RGBA or RGB depending on file)

    - other: uint8  [0,255]

    We normalise everything to float32 [0,1] and strip the alpha channel.

    """
    img = plt.imread(path)
    # Normalise uint8 to float
    if img.dtype == np.uint8:
        img = img.astype(np.float32) / 255.0
    # Drop alpha channel if present
    if img.ndim == 3 and img.shape[2] == 4:
        img = img[:, :, :3]
    # Clip to valid range (handles tiny float rounding errors)
    img = np.clip(img, 0.0, 1.0)
    if flip_h:
        img = img[:, ::-1, :]
    return img


# ---------------------------------------------------------------------------
# Core plotting function
# ---------------------------------------------------------------------------


def plot_sample(

    dataset: str,

    sample_id: str,

    out_dir: Optional[str] = None,

    dpi: int = 300,

    verbose: bool = True,

) -> Optional[str]:
    """Generate a 4×4 comparison grid for *sample_id* in *dataset*.



    Returns the path of the saved figure, or None on failure.

    """
    data_dir = os.path.join(ROOT, "data", dataset)
    results_dir = os.path.join(ROOT, "results", dataset)
    inputs_dir = os.path.join(data_dir, "inputs")
    gt_dir = os.path.join(data_dir, "GT")

    # ---- Locate source files -----------------------------------------------
    a1 = _find_input(inputs_dir, sample_id, "A1")
    a2 = _find_input(inputs_dir, sample_id, "A2")
    a3 = _find_input(inputs_dir, sample_id, "A3")
    gt = os.path.join(gt_dir, f"{sample_id}.png")

    missing: list[str] = []
    for tag, path in [("A1", a1), ("A2", a2), ("A3", a3), ("GT", gt)]:
        if not path or not os.path.exists(path):
            missing.append(tag)

    if missing:
        print(f"[WARN] {dataset}/{sample_id}: missing {missing} – skipping.")
        return None

    # ---- Build image grid --------------------------------------------------
    flip_set = FLIP_H_FOR_DISPLAY.get(dataset, set())

    grid: list[np.ndarray] = [
        _load(a1),
        _load(a2),
        _load(a3),
        _load(gt),
    ]

    for method in METHODS:
        pred_path = os.path.join(results_dir, method, f"{sample_id}.png")
        flip = method in flip_set
        if os.path.exists(pred_path):
            grid.append(_load(pred_path, flip_h=flip))
        else:
            if verbose:
                print(
                    f"  [WARN] missing {dataset}/{method}/{sample_id}.png  → black panel"
                )
            # Placeholder: black image with same shape as GT
            grid.append(np.zeros_like(grid[3]))

    assert len(grid) == 16, f"Expected 16 panels, got {len(grid)}"

    # ---- Render figure -----------------------------------------------------
    fig, axes = plt.subplots(4, 4, figsize=(8, 8), dpi=dpi)
    fig.subplots_adjust(
        left=0.01,
        right=0.99,
        top=0.99,
        bottom=0.06,
        wspace=0.04,
        hspace=0.10,
    )

    for idx, (ax, img, label) in enumerate(zip(axes.flat, grid, ALL_LABELS)):
        ax.imshow(img)
        ax.set_title(label, y=-0.18, fontsize=7)
        ax.axis("off")

    # ---- Save --------------------------------------------------------------
    if out_dir is None:
        out_dir = os.path.join(ROOT, "eval", "plots")
    os.makedirs(out_dir, exist_ok=True)

    out_path = os.path.join(out_dir, f"{dataset}_{sample_id}.pdf")
    fig.savefig(out_path, bbox_inches="tight")
    plt.close(fig)

    if verbose:
        print(f"Saved: {out_path}")
    return out_path


# ---------------------------------------------------------------------------
# Batch helpers
# ---------------------------------------------------------------------------


def available_ids(dataset: str) -> list[str]:
    """Return sorted list of sample IDs that have at least one input image."""
    inputs_dir = os.path.join(ROOT, "data", dataset, "inputs")
    a1_files = sorted(glob(os.path.join(inputs_dir, "*_A1.png")))
    return [os.path.basename(f).replace("_A1.png", "") for f in a1_files]


def generate_paper_figures(

    datasets: Optional[list[str]] = None,

    out_dir: Optional[str] = None,

) -> list[str]:
    """Generate all figures referenced in the paper."""
    if datasets is None:
        datasets = DATASETS
    saved: list[str] = []
    for ds in datasets:
        for sid in PAPER_SAMPLES.get(ds, []):
            print(f"\n--- {ds} / {sid} ---")
            path = plot_sample(ds, sid, out_dir=out_dir)
            if path:
                saved.append(path)
    return saved


# ---------------------------------------------------------------------------
# CLI
# ---------------------------------------------------------------------------


def _parse_args() -> argparse.Namespace:
    p = argparse.ArgumentParser(
        description="Generate comparison figures for cloud-removal methods"
    )
    p.add_argument(
        "--dataset",
        type=str,
        default=None,
        choices=DATASETS,
        help="Dataset to use (default: both when --paper-samples is set)",
    )
    p.add_argument(
        "--id",
        type=str,
        default=None,
        metavar="SAMPLE_ID",
        help="Generate a figure for this specific sample ID",
    )
    p.add_argument(
        "--paper-samples",
        action="store_true",
        help="Generate the exact figures used in the paper",
    )
    p.add_argument(
        "--all",
        action="store_true",
        help="Generate figures for ALL available samples in the chosen dataset",
    )
    p.add_argument(
        "--list",
        action="store_true",
        help="List available sample IDs and exit",
    )
    p.add_argument(
        "--out-dir",
        type=str,
        default=None,
        help="Output directory (default: eval/plots/)",
    )
    p.add_argument(
        "--dpi",
        type=int,
        default=300,
        help="Figure resolution in DPI (default: 300)",
    )
    return p.parse_args()


def main() -> None:
    args = _parse_args()

    # Determine which datasets to process
    if args.dataset:
        datasets = [args.dataset]
    else:
        datasets = DATASETS

    # ---- list mode ---------------------------------------------------------
    if args.list:
        for ds in datasets:
            ids = available_ids(ds)
            print(f"\n{ds}  ({len(ids)} samples)")
            for i, sid in enumerate(ids):
                print(f"  {sid}")
                if i >= 29 and len(ids) > 30:
                    print(f"  ... and {len(ids) - 30} more (use --all to see all)")
                    break
        return

    # ---- paper figures -----------------------------------------------------
    if args.paper_samples:
        saved = generate_paper_figures(datasets=datasets, out_dir=args.out_dir)
        print(f"\n{len(saved)} figure(s) saved.")
        return

    # ---- single sample -----------------------------------------------------
    if args.id:
        if len(datasets) > 1:
            print("[INFO] --id specified without --dataset; trying both datasets.")
        for ds in datasets:
            plot_sample(ds, args.id, out_dir=args.out_dir, dpi=args.dpi)
        return

    # ---- all samples -------------------------------------------------------
    if args.all:
        if not args.dataset:
            print("[ERROR] Please specify --dataset when using --all.")
            return
        ids = available_ids(args.dataset)
        print(f"Generating {len(ids)} figures for {args.dataset} …")
        for sid in ids:
            plot_sample(
                args.dataset, sid, out_dir=args.out_dir, dpi=args.dpi, verbose=False
            )
            print(f"  done: {sid}")
        print("Finished.")
        return

    # ---- no action specified -----------------------------------------------
    print(
        "No action specified. Examples:\n"
        "  python plot.py --paper-samples\n"
        "  python plot.py --dataset Sen2_MTC_New --id T12TUR_R027_55\n"
        "  python plot.py --dataset Sen2_MTC_New --list\n"
        "  python plot.py --dataset Sen2_MTC_New --all\n"
    )


if __name__ == "__main__":
    main()