Add hero gallery figure

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

scripts/make_gallery.py

@@ -0,0 +1,88 @@
+"""Render the hero gallery: 4 sample images x (raw | cluster masks | berry keypoints)."""
+import json
+import sys
+from pathlib import Path
+
+import matplotlib
+
+matplotlib.use("Agg")
+import matplotlib.pyplot as plt
+import numpy as np
+import pandas as pd
+from matplotlib.patches import Polygon
+from PIL import Image
+
+sys.path.insert(0, str(Path(__file__).resolve().parent))
+from vivid_utils import STYLE
+
+ROOT = Path(__file__).resolve().parents[1]
+DATA = ROOT / "data"
+GRAPE_CATEGORY_ID = 1
+MASK_COLORS = ["#6366F1", "#F59E0B", "#10B981", "#EF4444", "#8B5CF6", "#06B6D4"]
+
+# Override with specific file_names after eyeballing candidates; None = auto-pick
+GALLERY_FILES = None
+
+
+def pick_samples(meta):
+    if GALLERY_FILES:
+        return list(GALLERY_FILES)
+    picks = []
+    pool = meta[meta.cluster_count.between(2, 8)].sort_values("berry_count")
+    for q in (0.15, 0.45, 0.7, 0.95):
+        row = pool.iloc[min(int(q * len(pool)), len(pool) - 1)]
+        if row.file_name not in picks:
+            picks.append(row.file_name)
+    return picks
+
+
+def main():
+    meta = pd.read_csv(DATA / "metadata.csv")
+    samples = pick_samples(meta)
+    print("gallery samples:", samples)
+
+    print("Loading COCO json...")
+    coco = json.loads((DATA / "anns" / "instances_updated.json").read_text())
+    ids = {im["file_name"]: im["id"] for im in coco["images"] if im["file_name"] in samples}
+    polys = {name: [] for name in samples}
+    id_to_name = {v: k for k, v in ids.items()}
+    for ann in coco["annotations"]:
+        if ann["category_id"] != GRAPE_CATEGORY_ID or ann["image_id"] not in id_to_name:
+            continue
+        for seg in ann["segmentation"]:
+            pts = np.asarray(seg, dtype=float).reshape(-1, 2)
+            polys[id_to_name[ann["image_id"]]].append(pts)
+
+    fig, axes = plt.subplots(len(samples), 3, figsize=(9, 3.9 * len(samples)),
+                             dpi=STYLE["dpi"], facecolor="white")
+    col_titles = ["image", "cluster masks", "berry keypoints"]
+    for r, name in enumerate(samples):
+        img = Image.open(DATA / "imgs" / name).convert("RGB")
+        img.thumbnail((900, 900))
+        scale = img.width / meta.set_index("file_name").loc[name, "width"]
+        pts = np.load(DATA / "anns" / "points" / (Path(name).stem + ".npy")) * scale
+        for c in range(3):
+            ax = axes[r, c]
+            ax.imshow(img)
+            ax.set_xticks([]), ax.set_yticks([])
+            for spine in ax.spines.values():
+                spine.set_visible(False)
+            if r == 0:
+                ax.set_title(col_titles[c], fontsize=12, color=STYLE["text"])
+        for i, poly in enumerate(polys[name]):
+            axes[r, 1].add_patch(Polygon(poly * scale, closed=True, alpha=0.45,
+                                         facecolor=MASK_COLORS[i % len(MASK_COLORS)],
+                                         edgecolor="white", linewidth=1.0))
+        axes[r, 2].scatter(pts[:, 0], pts[:, 1], s=4, color="#FDE047",
+                           edgecolors="#B45309", linewidths=0.3)
+        axes[r, 2].text(0.02, 0.02, f"{len(pts)} berries", transform=axes[r, 2].transAxes,
+                        fontsize=11, fontweight="bold", color="white",
+                        bbox=dict(facecolor="black", alpha=0.6, pad=3))
+    fig.tight_layout()
+    out = ROOT / "assets" / "hero_gallery.png"
+    fig.savefig(out, dpi=STYLE["dpi"], facecolor="white", bbox_inches="tight")
+    print("wrote", out)
+
+
+if __name__ == "__main__":
+    main()