Syn4D: A Multiview Synthetic 4D Dataset

Syn4D is a synthetic 4D dataset with multi-view RGB videos, depth, masks, tracking geometry, and supporting object mesh metadata.

Layout

data/
  syn4d_v1_stride_1/       # Syn4D V1, every frame can be a tracking reference frame
  syn4d_v1_stride_5/       # Syn4D V1, every 5th frame can be a tracking reference frame
  syn4d_sim/               # Kubric-style simulation subset
  metadata/new_weight_bone/

code/visualizer/
  README.md
  requirements.txt
  run.py

code/bedlam2/
  README.md
  smplx_anim_to_alembic_batch.py
  smplx_anim_to_objs_batch.py

LICENSE
README.md
THIRD_PARTY_NOTICES.md
LICENSES/
resources/

Each Syn4D V1 split includes sequence_to_asset_mapping.csv, which maps dataset sequences to their underlying assets. The same mapping files are also referenced in THIRD_PARTY_NOTICES.md together with third-party asset notices and license references.

Using The Dataset

1. Clone or download this Hugging Face dataset repository.

git lfs install
git clone https://huggingface.co/datasets/Syn4D/Syn4D
cd Syn4D

2. Choose a dataset subset and extract its scene archives.

The main subsets are:

• data/syn4d_v1_stride_1/: Syn4D V1 with every frame available as a possible tracking reference frame.
• data/syn4d_v1_stride_5/: Syn4D V1 with every 5th frame available as a possible tracking reference frame.
• data/syn4d_sim/: additional Kubric-style simulation subset packaged in the same safetensor/MP4 layout.

To extract all scene archives directly under the same subset folder:

SUBSET=data/syn4d_v1_stride_5

for archive in "$SUBSET"/*.tar.zst; do
  tar -I zstd -xf "$archive" -C "$SUBSET"
done

Equivalent decompression form if your tar does not support -I zstd:

tar --use-compress-program=unzstd -xf <archive>.tar.zst -C <output_dir>

For full geometry/body reconstruction workflows, use the metadata under data/metadata/. Object metadata is provided under data/metadata/new_weight_bone/. Processed BEDLAM2 metadata is not redistributed due to BEDLAM2 license restrictions; if needed, follow code/bedlam2/README.md to download BEDLAM2 from the official source and convert SMPL-X motion files.

BEDLAM2 conversion is CPU-heavy, so running it on machines with fast CPUs is recommended.

3. Run the visualizer or integrate the dataset.

For visualization instructions, see code/visualizer/README.md.