Add scripts/preprocess_step2_qc_plots.py

scripts/preprocess_step2_qc_plots.py

@@ -0,0 +1,152 @@
+#!/usr/bin/env python3
+"""
+preprocess_step2_qc_plots.py
+NEST3D QC plots for visual inspection of PLY files.
+
+For each sample, generates a 3-by-2 grid:
+  Row 1: Top view    (X,Y) - labels | RGB
+  Row 2: Side view   (X,Z) - labels | RGB
+  Row 3: Front view  (Y,Z) - labels | RGB
+Stats box shows point counts per class.
+
+Usage:
+  python preprocess_step2_qc_plots.py --data-dir /path/to/reconstructions --version original
+  python preprocess_step2_qc_plots.py --data-dir /path/to/reconstructions --version corrected
+  python preprocess_step2_qc_plots.py --data-dir /path/to/reconstructions --version corrected --samples sample001 sample002
+
+Output: <data-dir>/sampleXXX/sampleXXX_qc_{version}.png
+"""
+
+import argparse
+import numpy as np
+import matplotlib
+matplotlib.use("Agg")
+import matplotlib.pyplot as plt
+import matplotlib.patches as mpatches
+from plyfile import PlyData
+from pathlib import Path
+
+MAX_PLOT_PTS = 150_000
+
+CLASS_COLORS = {
+    0:   np.array([0.2, 0.7, 0.2]),
+    1:   np.array([0.6, 0.3, 0.1]),
+    2:   np.array([0.9, 0.1, 0.1]),
+    255: np.array([0.7, 0.7, 0.7]),
+}
+LABEL_NAMES = {0:"grass(0)", 1:"tree(1)", 2:"nest(2)", 255:"ignore(255)"}
+
+def load_ply(path):
+    ply = PlyData.read(str(path))
+    v = ply["vertex"]
+    xyz = np.stack([v["x"],v["y"],v["z"]], axis=1).astype(np.float32)
+    rgb = np.stack([v["red"],v["green"],v["blue"]], axis=1).astype(np.float32)/255.0
+    lbl = np.array(v["scalar_Classification"], dtype=np.int32)
+    return xyz, rgb, lbl
+
+def subsample(xyz, rgb, lbl, n=MAX_PLOT_PTS):
+    if len(xyz) <= n:
+        return xyz, rgb, lbl
+    idx = np.random.default_rng(42).choice(len(xyz), n, replace=False)
+    return xyz[idx], rgb[idx], lbl[idx]
+
+def make_label_colors(lbl):
+    c = np.zeros((len(lbl),3), np.float32)
+    for k,col in CLASS_COLORS.items():
+        c[lbl==k] = col
+    return c
+
+def scatter2d(ax, a, b, colors, s, title, xl, yl):
+    ax.scatter(a, b, c=colors, s=s, linewidths=0, rasterized=True)
+    ax.set_title(title, fontsize=8, pad=3)
+    ax.set_xlabel(xl, fontsize=7)
+    ax.set_ylabel(yl, fontsize=7)
+    ax.tick_params(labelsize=6)
+    ax.set_aspect("equal")
+
+def make_plot(sample_id, ply_path, version):
+    out_png = ply_path.parent / f"{sample_id}_qc_{version}.png"
+    if out_png.exists():
+        print(f"[SKIP] {sample_id} ({version})")
+        return
+
+    print(f"[PLOT] {sample_id} ({version}) ...", end=" ", flush=True)
+    try:
+        xyz, rgb, lbl = load_ply(ply_path)
+    except Exception as e:
+        print(f"ERROR: {e}")
+        return
+
+    total = len(lbl)
+    xyz_s, rgb_s, lbl_s = subsample(xyz, rgb, lbl)
+    lbl_colors = make_label_colors(lbl_s)
+    dot = max(0.2, min(1.5, 80_000/len(xyz_s)))
+
+    unique, counts = np.unique(lbl, return_counts=True)
+    stats = dict(zip(unique.tolist(), counts.tolist()))
+    nest_pct = 100*stats.get(2,0)/total if total>0 else 0
+
+    stats_lines = [f"Total: {total:,}", ""]
+    for k in [0,1,2,255]:
+        c = stats.get(k,0)
+        stats_lines.append(f"{LABEL_NAMES[k]}: {c:,} ({100*c/total:.1f}%)")
+    stats_lines += ["", f"Nest ~{nest_pct:.2f}%"]
+    stats_lines.append(f"X extent: {xyz[:,0].max()-xyz[:,0].min():.1f}m")
+    stats_lines.append(f"Y extent: {xyz[:,1].max()-xyz[:,1].min():.1f}m")
+    stats_lines.append(f"Z extent: {xyz[:,2].max()-xyz[:,2].min():.1f}m")
+
+    fig, axes = plt.subplots(3, 2, figsize=(12, 15))
+    fig.suptitle(f"{sample_id} [{version}]", fontsize=14, fontweight="bold", y=0.98)
+
+    views = [
+        (0,1,"X (m)","Y (m)","Top view"),
+        (0,2,"X (m)","Z (m)","Side view"),
+        (1,2,"Y (m)","Z (m)","Front view"),
+    ]
+    for row,(hi,vi,xl,yl,vname) in enumerate(views):
+        scatter2d(axes[row,0], xyz_s[:,hi], xyz_s[:,vi], lbl_colors, dot, f"{vname} - labels", xl, yl)
+        scatter2d(axes[row,1], xyz_s[:,hi], xyz_s[:,vi], rgb_s,      dot, f"{vname} - RGB",    xl, yl)
+
+    axes[2,1].text(0.98, 0.02, "\n".join(stats_lines),
+        transform=axes[2,1].transAxes, fontsize=6.5,
+        va="bottom", ha="right", family="monospace",
+        bbox=dict(boxstyle="round,pad=0.4", facecolor="white", alpha=0.75, edgecolor="gray"))
+
+    patches = [mpatches.Patch(color=CLASS_COLORS[k], label=LABEL_NAMES[k]) for k in [0,1,2,255]]
+    fig.legend(handles=patches, loc="lower center", ncol=4, fontsize=8, bbox_to_anchor=(0.5,0.01))
+
+    plt.tight_layout(rect=[0,0.03,1,0.97])
+    fig.savefig(str(out_png), dpi=100, bbox_inches="tight")
+    plt.close(fig)
+    print("saved")
+
+def main():
+    parser = argparse.ArgumentParser(description="NEST3D QC plots")
+    parser.add_argument(
+        "--data-dir", type=Path, default=Path("./reconstructions"),
+        help="Path to the reconstructions/ folder containing sampleXXX subfolders (default: ./reconstructions)"
+    )
+    parser.add_argument("--version", choices=["original","corrected"], default="corrected")
+    parser.add_argument("--samples", nargs="+", default=None)
+    args = parser.parse_args()
+    recon_dir = args.data_dir
+
+    sample_dirs = sorted(recon_dir.glob("sample*"))
+    if args.samples:
+        sample_dirs = [d for d in sample_dirs if d.name in args.samples]
+
+    print(f"Plotting {len(sample_dirs)} samples [{args.version}]\n")
+
+    for sample_dir in sample_dirs:
+        sample_id = sample_dir.name
+        suffix = "_corrected" if args.version=="corrected" else ""
+        ply_path = sample_dir / f"{sample_id}{suffix}.ply"
+        if not ply_path.exists():
+            print(f"[SKIP] {sample_id}: {ply_path.name} not found")
+            continue
+        make_plot(sample_id, ply_path, args.version)
+
+    print("\nAll done!")
+
+if __name__ == "__main__":
+    main()