| """Load aligned ZED RGB and metric depth without the ZED SDK.""" |
|
|
| from typing import Iterable, Tuple |
|
|
| import cv2 |
| import h5py |
| import numpy as np |
| from tqdm import tqdm |
|
|
|
|
| def extract_camera_data( |
| rgb_video_path: str, |
| depth_h5_path: str, |
| frame_indices: Iterable[int], |
| ) -> Tuple[np.ndarray, np.ndarray, np.ndarray]: |
| """Return selected rectified-left RGB frames and aligned uint16-mm depth. |
| |
| Frame index zero is the first frame in both files. Depth zero means invalid. |
| """ |
| indices = np.unique(np.asarray(list(frame_indices), dtype=np.int64)) |
| if indices.size == 0: |
| return (np.empty((0, 0, 0, 3), dtype=np.uint8), |
| np.empty((0, 0, 0), dtype=np.uint16), |
| np.empty(4, dtype=np.float32)) |
| if indices[0] < 0: |
| raise ValueError("ZED frame indices must be nonnegative") |
|
|
| with h5py.File(depth_h5_path, "r") as f: |
| if "depth_mm" not in f: |
| raise ValueError(f"Missing depth_mm dataset: {depth_h5_path}") |
| depth = f["depth_mm"] |
| if depth.ndim != 3 or depth.dtype != np.uint16: |
| raise ValueError("depth_mm must have shape (frames, height, width) and dtype uint16") |
| frame_count, height, width = depth.shape |
| if indices[-1] >= frame_count: |
| raise ValueError(f"ZED index {indices[-1]} exceeds {frame_count} depth frames") |
| intrinsics = np.asarray(f.attrs["intrinsics_fx_fy_cx_cy"], dtype=np.float32) |
| if intrinsics.shape != (4,): |
| raise ValueError("Invalid ZED intrinsics") |
|
|
| cap = cv2.VideoCapture(rgb_video_path) |
| if not cap.isOpened(): |
| raise RuntimeError(f"Could not open ZED RGB video: {rgb_video_path}") |
| try: |
| video_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT)) |
| if video_count > 0 and video_count != frame_count: |
| raise ValueError(f"ZED RGB/depth frame count differs: {video_count} vs {frame_count}") |
| if (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)), |
| int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))) != (width, height): |
| raise ValueError("ZED RGB/depth resolution differs") |
| rgb_frames = [] |
| depth_frames = [] |
| wanted = set(indices.tolist()) |
| for frame_idx in tqdm(range(int(indices[-1]) + 1), desc="ZED extraction", unit="frame"): |
| ok, bgr = cap.read() |
| if not ok: |
| raise ValueError(f"ZED RGB video ends before frame {frame_idx}") |
| if frame_idx in wanted: |
| rgb_frames.append(cv2.cvtColor(bgr, cv2.COLOR_BGR2RGB)) |
| depth_frames.append(depth[frame_idx]) |
| finally: |
| cap.release() |
|
|
| return np.stack(rgb_frames), np.stack(depth_frames), intrinsics |
|
|
