README.md

metadata

language:
  - en
tags:
  - motion
  - smpl
  - stickman
size_categories:
  - 1M<n<10M

motion_example_0801

This dataset contains 2,000,000 motion examples distributed as 1,954 uncompressed tar archives, totaling 1,277,023,825,920 bytes (approximately 1.28 TB).

Each example has three corresponding files:

Directory	Format	Contents
human_motion/	.pt	SMPL motion data
motion_caption/	.json	Motion descriptions
rendered_videos/	.mp4	Stickman renderings of the motion

Archives are named motion_example_0801_00000.tar through motion_example_0801_01953.tar. The first 1,953 archives contain 1,024 examples each; the last contains 128. archive_index.tsv lists the example count for each archive.

The first archive also contains these five metadata files at its beginning:

• metadata.jsonl
• metadata_caption.jsonl
• metadata_classification.jsonl
• metadata_objects.jsonl
• metadata_render.jsonl

In metadata_caption.jsonl, each video_caption combines descriptions of its two source motion segments. One of eight transition templates is randomly selected, with gender-neutral person references. These captions retain descriptions of the source videos, including their appearance and setting, and are not literal descriptions of the stickman renderings. The video_path field is empty; rendering information is provided in metadata_render.jsonl.

Verify and extract

Download all archive shards together with SHA256SUMS and archive_index.tsv. Verify the archives before extraction:

sha256sum -c SHA256SUMS

Extract every archive into the same directory. Archive member paths are relative to the dataset root and do not include an additional enclosing directory:

mkdir -p motion_example_0801
for archive in motion_example_0801_*.tar; do
    tar -xf "$archive" -C motion_example_0801
done

The extracted directory contains the five metadata files and the three data directories listed above. The tar files are uncompressed, so retaining both the archives and extracted dataset requires space for both copies.