File size: 6,251 Bytes
0e4e948
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
"""
Batch mesh-mapper: assign mesh faces to fitted superquadrics (reuses existing
SQ-fit + curvature atoms).

Inputs (all via CLI):
  --mesh_dir     <mesh_dir>/<uid>/full.ply
  --v4_dir       <v4_dir>/<uid>/face_labels.npy        (curvature atoms)
  --sq_root      <sq_root>/<uid>/sq_fit_v20/post_sq_*.ply
  --output_root  destination

Per UID:
  1. load mesh / face_labels / post_sq_*.ply
  2. run mesh_mapper.map_with_atoms
  3. write face_labels_v8.npy + mesh_mapped_v8.ply + report.json
  4. aggregate to batch_summary.csv / .json

Supports --skip_existing for resume.
"""
from __future__ import annotations

import os
import sys
import csv
import json
import time
import glob
import argparse
import numpy as np
import trimesh
from tqdm import tqdm

from hitops.mapping.mesh_mapper import (
    _load_sq_voxel_groups,
    map_with_atoms,
    save_results,
)


def discover_uids(v4_dir: str) -> list[str]:
    uids = []
    for d in sorted(os.listdir(v4_dir)):
        full = os.path.join(v4_dir, d)
        if os.path.isdir(full) and os.path.isfile(os.path.join(full, "face_labels.npy")):
            uids.append(d)
    return uids


def process_one(
    uid: str, mesh_dir: str, v4_dir: str, sq_root: str, output_root: str,
    *, vote_mode="count", vote_tau=0.025,
    orphan_face_dist=0.04, orphan_atom_frac=0.6, min_atom_size=3,
    save_per_sq=False,
) -> dict:
    mesh_path = os.path.join(mesh_dir, uid, "full.ply")
    fl_path   = os.path.join(v4_dir, uid, "face_labels.npy")
    sq_dir    = os.path.join(sq_root, uid, "sq_fit_v20")
    out_dir   = os.path.join(output_root, uid)

    for p, name in [(mesh_path, "mesh"), (fl_path, "face_labels"), (sq_dir, "sq_dir")]:
        if not os.path.exists(p):
            return {"uid": uid, "status": "skipped", "reason": f"missing {name}: {p}"}

    t0 = time.time()
    mesh = trimesh.load(mesh_path, force="mesh", process=False)
    face_labels_v4 = np.load(fl_path).astype(np.int32)
    sq_names, sq_pts = _load_sq_voxel_groups(sq_dir)
    if not sq_names:
        return {"uid": uid, "status": "skipped", "reason": "empty sq_dir"}

    res = map_with_atoms(
        mesh, face_labels_v4, sq_pts,
        vote_mode=vote_mode, vote_tau=vote_tau,
        orphan_face_dist=orphan_face_dist,
        orphan_atom_frac=orphan_atom_frac,
        min_atom_size=min_atom_size,
    )
    save_results(mesh, res, sq_names, out_dir, save_per_sq=save_per_sq)
    t_total = time.time() - t0

    s = {
        "uid": uid, "status": "ok",
        "t_total": round(t_total, 2),
        **res["stats"],
    }
    return s


def main():
    ap = argparse.ArgumentParser(description="HY3D batch runner for mesh_mapper_v8")
    ap.add_argument("--mesh_dir",    type=str, required=True,
                    help="mesh dir; meshes at <mesh_dir>/<uid>/full.ply")
    ap.add_argument("--v4_dir",      type=str, required=True,
                    help="curvature-seg dir; labels at <v4_dir>/<uid>/face_labels.npy")
    ap.add_argument("--sq_root",     type=str, required=True,
                    help="SQ-fit root; SQs at <sq_root>/<uid>/sq_fit_v20/")
    ap.add_argument("--output_root", type=str, required=True)
    ap.add_argument("--vote_mode",   type=str, default="count", choices=["count", "exp"])
    ap.add_argument("--vote_tau",    type=float, default=0.025)
    ap.add_argument("--orphan_face_dist", type=float, default=0.04)
    ap.add_argument("--orphan_atom_frac", type=float, default=0.6)
    ap.add_argument("--min_atom_size",    type=int, default=3)
    ap.add_argument("--save_per_sq",      action="store_true")
    ap.add_argument("--skip_existing",    action="store_true",
                    help="skip a UID if output/<uid>/face_labels_v8.npy already exists")
    args = ap.parse_args()

    uids = discover_uids(args.v4_dir)
    print(f"[batch-v8] {len(uids)} UID(s) to process", flush=True)
    os.makedirs(args.output_root, exist_ok=True)

    stats: list[dict] = []
    for uid in tqdm(uids, desc="mapper_v8", ncols=100):
        out_dir = os.path.join(args.output_root, uid)
        if args.skip_existing and os.path.exists(os.path.join(out_dir, "face_labels_v8.npy")):
            stats.append({"uid": uid, "status": "skipped", "reason": "already_done"})
            continue
        try:
            s = process_one(
                uid, args.mesh_dir, args.v4_dir, args.sq_root, args.output_root,
                vote_mode=args.vote_mode, vote_tau=args.vote_tau,
                orphan_face_dist=args.orphan_face_dist,
                orphan_atom_frac=args.orphan_atom_frac,
                min_atom_size=args.min_atom_size,
                save_per_sq=args.save_per_sq,
            )
        except Exception as e:
            import traceback
            traceback.print_exc()
            s = {"uid": uid, "status": "failed", "error": str(e)}
        stats.append(s)

        # Write incremental JSON summary after each UID.
        with open(os.path.join(args.output_root, "batch_summary.json"), "w") as f:
            json.dump(stats, f, indent=2, default=lambda x: float(x) if isinstance(x, np.floating) else int(x))

    # csv
    csv_path = os.path.join(args.output_root, "batch_summary.csv")
    keys_order = [
        "uid", "status", "t_total", "F", "A", "K",
        "n_face_orphan_NN", "n_v4_orphan_face", "n_atom_orphan",
        "atom_size_min", "atom_size_median", "atom_size_max",
        "atom_conf_mean", "atom_conf_p25",
        "face_coverage", "sqs_used",
    ]
    with open(csv_path, "w", newline="") as f:
        w = csv.DictWriter(f, fieldnames=keys_order, extrasaction="ignore")
        w.writeheader()
        for s in stats:
            w.writerow({k: s.get(k, "") for k in keys_order})

    ok = [s for s in stats if s.get("status") == "ok"]
    print(f"\n[batch-v8] {len(ok)}/{len(stats)} OK  | csv={csv_path}")
    if ok:
        print(f"  t_total:        mean={np.mean([s['t_total']    for s in ok]):.2f}s")
        print(f"  face_coverage:  mean={np.mean([s['face_coverage'] for s in ok])*100:.2f}%")
        print(f"  sqs_used/K:     "
              f"mean={np.mean([s['sqs_used']/max(s['K'],1) for s in ok])*100:.1f}%")
        print(f"  atom_conf_mean: mean={np.mean([s['atom_conf_mean'] for s in ok]):.3f}")


if __name__ == "__main__":
    main()