ZhengGuangze/DynamicReplica_vlbm

DynamicReplica (converted to VLBM format)

This dataset contains DynamicReplica sequences converted to the VLBM/Flock4D-compatible format using the conversion tool dynamicreplica2vlbm.py.

Only the left camera view is extracted in this conversion.

Dataset Structure

Each sequence directory follows this layout:

{sequence_id}/
├── rgbs/
│   ├── rgb_00000.jpg
│   ├── rgb_00001.jpg
│   └── ...
├── depths/
│   ├── depth_00000.npz
│   ├── depth_00001.npz
│   └── ...
├── annotations.npz
└── scene_info.json

File Descriptions

• rgbs/: RGB images saved as JPEG (rgb_XXXXX.jpg). Images are converted to standard 3-channel RGB. Resolution may vary per sequence depending on the original data.

• depths/: Depth maps saved as compressed NumPy archives (depth_XXXXX.npz). Depth arrays were written using np.savez_compressed and stored in float16 for compactness. Depth values are the metric depth as provided by the DynamicReplica source (units follow the original dataset).

• annotations.npz: NumPy compressed file containing the following arrays (types are float16 where noted):

  • trajs_2d: 2D trajectories (T, N, 2) — pixel coordinates (x, y). If trajectories are missing for the sequence, shape may be (T, 0, 2).
  • trajs_3d: 3D trajectories (T, N, 3) — 3D coordinates in world space (x, y, z).
  • visibilities: (T, N) — visibility flags (1.0 visible, 0.0 not visible).
  • intrinsics: (T, 3, 3) — camera intrinsic matrices for each frame.
  • extrinsics: (T, 4, 4) — world-to-camera extrinsic matrices (W2C) for each frame.

• scene_info.json: JSON file with per-sequence metadata written by the conversion script. Typical fields include source, sequence_name, num_frames, num_points, image_size, and camera.

Notes on Conversion

• The conversion was performed with dynamicreplica2vlbm.py. That script reads DynamicReplica frame annotations and per-frame trajectory files (when present), constructs PyTorch3D PerspectiveCameras from the dataset viewpoint metadata, and exports intrinsics and extrinsics using PyTorch3D's projection utilities.
• Only the left camera view is exported.
• Trajectory files in the original dataset may be absent for some frames or sequences; in that case the trajs_* arrays will be empty along the point dimension.
• Depth images in the original dataset are read from 16-bit PNGs and saved as .npz float16 arrays to reduce storage.

Data Specifications

• Image format: JPEG (RGB)
• Depth format: NPZ (float16)
• Annotation format: annotations.npz (float16 arrays for compact storage)
• Camera: left view only
• Frames per sequence: varies by sequence
• Points per sequence: varies; some sequences may contain zero tracked points after conversion

Usage Example (Python)

import numpy as np
from PIL import Image
from pathlib import Path
import json

seq_dir = Path("data/dynamicreplica_vlbm/0000")

# Load annotations
annotations = np.load(seq_dir / "annotations.npz", allow_pickle=True)
trajs_2d = annotations['trajs_2d']    # (T, N, 2)
trajs_3d = annotations['trajs_3d']    # (T, N, 3)
vis = annotations['visibilities']     # (T, N)
intrinsics = annotations['intrinsics']
extrinsics = annotations['extrinsics']

# Load an image and depth
frame_idx = 0
rgb = Image.open(seq_dir / "rgbs" / f"rgb_{frame_idx:05d}.jpg")
depth_npz = np.load(seq_dir / "depths" / f"depth_{frame_idx:05d}.npz")
depth = depth_npz['depth']  # float16 array (H, W)

# Load scene info
with open(seq_dir / "scene_info.json", 'r') as f:
    scene_info = json.load(f)

print(scene_info)

Citation

Please cite the original DynamicReplica dataset when using the converted data. Additionally, if you use the converted VLBM/Flock4D package, cite this repository and the conversion script.

Contact

If you encounter issues with the conversion or the converted files, please open an issue in the repository.