ZhengGuangze/DynamicReplica_vlbm

DynamicReplica (converted to VLBM format)

This dataset contains 484 sequences from the DynamicReplica dataset converted to the VLBM-compatible format using preprocess_dynamicreplica.py. The sequences have been compressed into .tar.gz archives in chunks of 50 sequences per archive.

Scale

Metric	Value
Total sequences	484
Image resolution	1280 x 720 px

Dataset Structure

Each sequence directory follows this layout:

<seq_name>/
├── rgbs/
│   ├── rgb_00000.jpg
│   ├── rgb_00001.jpg
│   └── ...
├── depths/
│   ├── depth_00000.npz
│   ├── depth_00001.npz
│   └── ...
├── intrinsics.npy
├── extrinsics.npy
├── trajs_2d.npy
├── trajs_3d.npy
├── visibilities.npy
└── scene_info.json

File Descriptions

• rgbs/: RGB frames from the left camera saved as JPEG (rgb_XXXXX.jpg). Resolution is 1280x720 pixels.
• depths/: Dense depth maps saved as compressed NumPy archives (depth_XXXXX.npz). Each archive stores a float16 array under the key depth of shape (H, W) in meters.
• intrinsics.npy: Camera intrinsic matrices for each frame (T, 3, 3) float16.
• extrinsics.npy: World-to-camera extrinsic matrices (W2C) for each frame (T, 4, 4) float16.
• trajs_2d.npy: 2D trajectories (T, N, 2) float16 -- pixel coordinates (x, y).
• trajs_3d.npy: 3D trajectories (T, N, 3) float16 -- world-space coordinates (x, y, z).
• visibilities.npy: Visibility flags (T, N) float16 (1.0 visible, 0.0 not visible).
• scene_info.json: JSON file with per-sequence metadata including source, sequence_name, num_frames, num_points, image_size, and camera.

Trajectory Source

DynamicReplica provides ground-truth dense point trajectories from its tracking annotations. Per-frame trajectory files contain 3D world coordinates and 2D projections with visibility flags.

Camera intrinsics and extrinsics are derived from PyTorch3D camera annotations using the official opencv_from_cameras_projection utility.

Usage Example (Python)

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

seq_dir = Path("data/dynamicreplica_vlbm/009850-3_obj")

# Load annotations
trajs_2d    = np.load(seq_dir / "trajs_2d.npy")     # (T, N, 2)
trajs_3d    = np.load(seq_dir / "trajs_3d.npy")     # (T, N, 3)
vis         = np.load(seq_dir / "visibilities.npy")  # (T, N)
intrinsics  = np.load(seq_dir / "intrinsics.npy")    # (T, 3, 3)
extrinsics  = np.load(seq_dir / "extrinsics.npy")    # (T, 4, 4)

# Load an image and depth map
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:

@inproceedings{karaev2023dynamicreplica,
  title={DynamicReplica: Learning a Dynamic Scene Representation on Real-World Video},
  author={Karaev, Nikita and Rocco, Ignacio and Graham, Benjamin and Neverova, Natalia and Vedaldi, Andrea and Rupprecht, Christian},
  booktitle={CVPR},
  year={2023}
}