File size: 9,878 Bytes
11c70d1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
"""Build ``points3d.ply`` from the first frame of every static camera.

The camera metadata uses Blender camera coordinates: +X right, +Y up, and
+Z backward. Depth therefore projects along -Z before the camera-to-world
transform is applied. ``CineCamera_Moving`` is intentionally excluded.
"""

from __future__ import annotations

import argparse
import json
from pathlib import Path, PurePosixPath
import re
import tempfile

import numpy as np
from PIL import Image


STATIC_CAMERA_METADATA = re.compile(r"^blender_(CineCamera_\d+)\.json$")


def static_camera_metadata(render_dir: Path) -> list[Path]:
    def camera_index(path: Path) -> int:
        match = STATIC_CAMERA_METADATA.fullmatch(path.name)
        if match is None:
            return -1
        return int(match.group(1).removeprefix("CineCamera_"))

    files = [
        path
        for path in render_dir.glob("blender_CineCamera_*.json")
        if STATIC_CAMERA_METADATA.fullmatch(path.name)
    ]
    return sorted(files, key=camera_index)


def first_frame_paths(render_dir: Path, frame: dict) -> tuple[Path, Path]:
    raw_path = str(frame["file_path"]).replace("\\", "/")
    relative = PurePosixPath(raw_path)
    parts = list(relative.parts)
    try:
        rgb_index = parts.index("rgb")
    except ValueError as exc:
        raise ValueError(f"Camera frame path does not contain an rgb component: {raw_path}") from exc

    rgb_relative = relative if relative.suffix.lower() == ".jpg" else relative.with_suffix(".jpg")
    parts[rgb_index] = "depth"
    depth_relative = PurePosixPath(*parts).with_suffix(".npz")
    return render_dir.joinpath(*rgb_relative.parts), render_dir.joinpath(*depth_relative.parts)


def load_rgb(path: Path) -> np.ndarray:
    with Image.open(path) as image:
        rgb = np.asarray(image.convert("RGB"))
    if rgb.ndim == 2:
        rgb = np.repeat(rgb[..., None], 3, axis=-1)
    if rgb.ndim != 3 or rgb.shape[2] < 3:
        raise ValueError(f"Unsupported RGB image shape for {path}: {rgb.shape}")
    rgb = rgb[..., :3]
    if rgb.dtype != np.uint8:
        if np.issubdtype(rgb.dtype, np.floating) and np.nanmax(rgb) <= 1.0:
            rgb = rgb * 255.0
        rgb = np.clip(rgb, 0, 255).astype(np.uint8)
    return rgb


def write_binary_ply(path: Path, points: np.ndarray, colors: np.ndarray) -> None:
    path.parent.mkdir(parents=True, exist_ok=True)
    vertex_type = np.dtype(
        [
            ("x", "<f8"),
            ("y", "<f8"),
            ("z", "<f8"),
            ("red", "u1"),
            ("green", "u1"),
            ("blue", "u1"),
        ]
    )
    vertices = np.empty(len(points), dtype=vertex_type)
    vertices["x"], vertices["y"], vertices["z"] = points.T
    vertices["red"], vertices["green"], vertices["blue"] = colors.T
    header = (
        "ply\n"
        "format binary_little_endian 1.0\n"
        "comment Created by PhysInOne Camera and Rendering Tools\n"
        f"element vertex {len(vertices)}\n"
        "property double x\n"
        "property double y\n"
        "property double z\n"
        "property uchar red\n"
        "property uchar green\n"
        "property uchar blue\n"
        "end_header\n"
    ).encode("ascii")

    with tempfile.NamedTemporaryFile(dir=path.parent, suffix=".ply", delete=False) as handle:
        temporary = Path(handle.name)
        handle.write(header)
        vertices.tofile(handle)
    try:
        if temporary.stat().st_size <= len(header):
            raise RuntimeError(f"PLY writer produced no vertices: {temporary}")
        temporary.chmod(0o664)
        temporary.replace(path)
    finally:
        temporary.unlink(missing_ok=True)


def valid_ply(path: Path) -> bool:
    if not path.is_file() or path.stat().st_size == 0:
        return False
    try:
        with path.open("rb") as handle:
            header = handle.read(2048).split(b"end_header\n", 1)[0].decode("ascii")
        match = re.search(r"^element vertex (\d+)$", header, flags=re.MULTILINE)
        return header.startswith("ply\n") and match is not None and int(match.group(1)) > 0
    except (OSError, UnicodeDecodeError, ValueError):
        return False


def generate_points3d(
    render_dir: Path,
    output: Path | None = None,
    depth_threshold: float = 20.0,
    max_points: int = 100_000,
    seed: int = 0,
) -> dict[str, int | float | str]:
    """Generate one point cloud using frame zero from every static view.

    Random priorities implement a bounded-memory uniform sample across all
    valid pixels from all views. The seed makes the output reproducible.
    """

    render_dir = render_dir.expanduser().resolve()
    output = render_dir / "points3d.ply" if output is None else output.expanduser().resolve()
    if depth_threshold <= 0:
        raise ValueError("depth_threshold must be positive")
    if max_points <= 0:
        raise ValueError("max_points must be positive")

    metadata_files = static_camera_metadata(render_dir)
    if not metadata_files:
        raise RuntimeError(f"No static blender_CineCamera_<index>.json files under {render_dir}")

    rng = np.random.default_rng(seed)
    kept_keys = np.empty(0, dtype=np.float64)
    kept_points = np.empty((0, 3), dtype=np.float64)
    kept_colors = np.empty((0, 3), dtype=np.uint8)
    valid_pixels = 0
    used_views = 0

    for metadata_path in metadata_files:
        with metadata_path.open("r", encoding="utf-8") as handle:
            document = json.load(handle)
        frames = document.get("frames", [])
        if not frames:
            raise RuntimeError(f"Camera metadata has no frames: {metadata_path}")

        frame = frames[0]
        rgb_path, depth_path = first_frame_paths(render_dir, frame)
        if not rgb_path.is_file() or not depth_path.is_file():
            raise FileNotFoundError(
                f"Missing first-frame input for {metadata_path.name}: rgb={rgb_path}, depth={depth_path}"
            )

        with np.load(depth_path) as depth_file:
            if "depth" not in depth_file:
                raise KeyError(f"Missing 'depth' array in {depth_path}")
            depth = np.asarray(depth_file["depth"], dtype=np.float64)
        rgb = load_rgb(rgb_path)
        if depth.ndim != 2 or rgb.shape[:2] != depth.shape:
            raise ValueError(
                f"RGB/depth shape mismatch for {metadata_path.name}: rgb={rgb.shape}, depth={depth.shape}"
            )

        height, width = depth.shape
        image_width = int(document.get("img_w", width))
        image_height = int(document.get("img_h", height))
        if (image_height, image_width) != (height, width):
            raise ValueError(
                f"Metadata/image shape mismatch for {metadata_path.name}: "
                f"metadata={(image_height, image_width)}, image={(height, width)}"
            )
        camera_angle_x = float(document["camera_angle_x"])
        focal = 0.5 * image_width / np.tan(0.5 * camera_angle_x)
        transform = np.asarray(frame["transform_matrix"], dtype=np.float64)
        if transform.shape != (4, 4):
            raise ValueError(f"Expected a 4x4 transform matrix in {metadata_path}")

        valid = np.isfinite(depth) & (depth > 0.0) & (depth < depth_threshold)
        flat_indices = np.flatnonzero(valid)
        if flat_indices.size == 0:
            raise RuntimeError(f"No valid first-frame depth pixels in {metadata_path.name}")
        valid_pixels += int(flat_indices.size)
        used_views += 1

        priorities = rng.random(flat_indices.size)
        if flat_indices.size > max_points:
            selected = np.argpartition(priorities, max_points - 1)[:max_points]
            flat_indices = flat_indices[selected]
            priorities = priorities[selected]

        rows, columns = np.divmod(flat_indices, width)
        z = depth.reshape(-1)[flat_indices]
        camera_points = np.column_stack(
            (
                (columns - image_width / 2.0) * z / focal,
                -(rows - image_height / 2.0) * z / focal,
                -z,
            )
        )
        world_points = camera_points @ transform[:3, :3].T + transform[:3, 3]
        colors = rgb.reshape(-1, 3)[flat_indices]

        kept_keys = np.concatenate((kept_keys, priorities))
        kept_points = np.concatenate((kept_points, world_points), axis=0)
        kept_colors = np.concatenate((kept_colors, colors), axis=0)
        if kept_keys.size > max_points:
            selected = np.argpartition(kept_keys, max_points - 1)[:max_points]
            kept_keys = kept_keys[selected]
            kept_points = kept_points[selected]
            kept_colors = kept_colors[selected]

    if used_views != len(metadata_files):
        raise RuntimeError(f"Used {used_views}/{len(metadata_files)} static views")
    write_binary_ply(output, kept_points, kept_colors)
    if not valid_ply(output):
        raise RuntimeError(f"Generated PLY did not pass validation: {output}")
    return {
        "views": used_views,
        "valid_pixels": valid_pixels,
        "points": int(len(kept_points)),
        "depth_threshold": depth_threshold,
        "output": str(output),
    }


def parse_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("render_dir", type=Path)
    parser.add_argument("--output", type=Path, default=None)
    parser.add_argument("--depth-threshold", type=float, default=20.0)
    parser.add_argument("--max-points", type=int, default=100_000)
    parser.add_argument("--seed", type=int, default=0)
    return parser.parse_args()


def main() -> int:
    args = parse_args()
    result = generate_points3d(
        args.render_dir,
        args.output,
        args.depth_threshold,
        args.max_points,
        args.seed,
    )
    print(json.dumps(result, indent=2, sort_keys=True))
    return 0


if __name__ == "__main__":
    raise SystemExit(main())