scripts/plot_mvsa_single.py

#!/usr/bin/env python3
"""Plot confusion matrix and ablation figure for MVSA-Single results."""

from __future__ import annotations

import argparse
import csv
from pathlib import Path
from typing import List, Tuple

import matplotlib

matplotlib.use("Agg")
import matplotlib.pyplot as plt
import numpy as np


def parse_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser(description="Plot MVSA-Single result figures")
    parser.add_argument("--results-dir", default="results/mvsa_single")
    parser.add_argument("--dpi", type=int, default=150)
    return parser.parse_args()


def read_confusion_matrix(path: Path) -> Tuple[List[str], np.ndarray]:
    with path.open("r", encoding="utf-8") as f:
        rows = list(csv.reader(f))

    if len(rows) < 2:
        raise ValueError(f"Invalid confusion matrix file: {path}")

    labels = [cell.strip() for cell in rows[0][1:] if cell.strip()]
    values = []
    for row in rows[1:]:
        values.append([int(float(x)) for x in row[1 : 1 + len(labels)]])

    cm = np.array(values, dtype=np.int64)
    return labels, cm


def read_ablation(path: Path) -> List[Tuple[str, float, float]]:
    rows: List[Tuple[str, float, float]] = []
    with path.open("r", encoding="utf-8") as f:
        reader = csv.DictReader(f)
        for row in reader:
            rows.append(
                (
                    str(row["variant"]),
                    float(row["accuracy"]),
                    float(row["f1_weighted"]),
                )
            )
    if not rows:
        raise ValueError(f"Empty ablation file: {path}")
    return rows


def plot_confusion_matrix(labels: List[str], cm: np.ndarray, out_path: Path, dpi: int) -> None:
    fig, ax = plt.subplots(figsize=(5.2, 4.4))

    row_sums = cm.sum(axis=1, keepdims=True).astype(np.float64)
    with np.errstate(divide="ignore", invalid="ignore"):
        cm_norm = np.divide(cm, row_sums, where=row_sums > 0)
        cm_norm = np.nan_to_num(cm_norm)

    im = ax.imshow(cm_norm, cmap="Blues", vmin=0.0, vmax=1.0)

    for i in range(cm.shape[0]):
        for j in range(cm.shape[1]):
            pct = cm_norm[i, j] * 100.0
            count = int(cm[i, j])
            color = "white" if cm_norm[i, j] > 0.55 else "black"
            ax.text(j, i, f"{count}\n({pct:.1f}%)", ha="center", va="center", fontsize=10, color=color)

    ax.set_xticks(np.arange(len(labels)))
    ax.set_yticks(np.arange(len(labels)))
    ax.set_xticklabels([label.capitalize() for label in labels], fontsize=10)
    ax.set_yticklabels([label.capitalize() for label in labels], fontsize=10)
    ax.set_xlabel("Predicted", fontsize=11)
    ax.set_ylabel("True", fontsize=11)

    cbar = fig.colorbar(im, ax=ax, fraction=0.046, pad=0.04)
    cbar.ax.tick_params(labelsize=9)

    fig.tight_layout()
    fig.savefig(out_path, dpi=dpi, bbox_inches="tight")
    plt.close(fig)


def plot_ablation_table(rows: List[Tuple[str, float, float]], out_path: Path, dpi: int) -> None:
    navy = "#2c3e6b"
    orange = "#d4a017"
    light_grey = "#f0f0f0"

    fig, ax = plt.subplots(figsize=(6.2, 3.4))
    ax.axis("off")

    col_labels = ["Variant", "Acc", "F1-Weighted"]
    table_data = [[name, f"{acc:.4f}", f"{f1w:.4f}"] for name, acc, f1w in rows]

    table = ax.table(
        cellText=table_data,
        colLabels=col_labels,
        loc="center",
        cellLoc="center",
    )
    table.auto_set_font_size(False)
    table.set_fontsize(10)
    table.scale(1.3, 1.8)

    for col_idx in range(len(col_labels)):
        cell = table[0, col_idx]
        cell.set_facecolor(navy)
        cell.set_text_props(color="white", fontweight="bold")
        cell.set_edgecolor("white")

    full_f1 = rows[0][2]
    for row_idx, (variant, _, f1w) in enumerate(rows, start=1):
        is_full = variant == "Full"
        for col_idx in range(len(col_labels)):
            cell = table[row_idx, col_idx]
            cell.set_edgecolor("#cccccc")
            if is_full:
                cell.set_facecolor("#dce6f1")
            elif row_idx % 2 == 0:
                cell.set_facecolor(light_grey)
        if (f1w > full_f1) and (not is_full):
            table[row_idx, 2].set_facecolor(orange)
            table[row_idx, 2].set_text_props(fontweight="bold")

    fig.tight_layout()
    fig.savefig(out_path, dpi=dpi, bbox_inches="tight")
    plt.close(fig)


def main() -> None:
    args = parse_args()
    results_dir = Path(args.results_dir)

    cm_path = results_dir / "confusion_matrix.csv"
    ablation_path = results_dir / "ablation_summary.csv"

    if not cm_path.exists():
        raise FileNotFoundError(f"Missing: {cm_path}")
    if not ablation_path.exists():
        raise FileNotFoundError(f"Missing: {ablation_path}")

    labels, cm = read_confusion_matrix(cm_path)
    ablation_rows = read_ablation(ablation_path)

    out_cm = results_dir / "figure4a_confusion_matrix.png"
    out_ablation = results_dir / "figure4b_ablation_study.png"

    plot_confusion_matrix(labels, cm, out_cm, args.dpi)
    plot_ablation_table(ablation_rows, out_ablation, args.dpi)

    print(f"Saved -> {out_cm}")
    print(f"Saved -> {out_ablation}")


if __name__ == "__main__":
    main()