SpaRRTa: A Synthetic Benchmark for Evaluating Spatial Intelligence in Visual Foundation Models

Summary

• Format: parquet shards with one row per sample.
• Split: train only.
• Images are embedded in parquet under the image struct column (bytes, path).
• Full raw metadata is preserved in raw_params_json.

Run Stats

• Discovered paired samples: 149145
• Skipped broken/missing pairs or parse errors: 0

Scene Coverage

scene_variant	scene	variant	paired_samples	new_rows	skipped_existing	missing_pairs
bridge	bridge	1	9834	0	9834	0
bridge_2	bridge	2	9834	0	9834	0
bridge_3	bridge	3	9834	0	9834	0
city	city	1	10000	0	10000	0
city_2	city	2	10000	0	10000	0
city_3	city	3	10000	0	10000	0
desert	desert	1	10000	0	10000	0
desert_2	desert	2	10000	0	10000	0
desert_3	desert	3	10000	0	10000	0
forest	forest	1	10000	0	10000	0
forest_2	forest	2	10000	0	10000	0
forest_3	forest	3	10000	0	10000	0
winter_town	winter_town	1	9881	0	9881	0
winter_town_2	winter_town	2	9881	0	9881	0
winter_town_3	winter_town	3	9881	0	9881	0

Columns

• sample_id (string): stable unique id (scene_variant:frame_id)
• scene (string): base scene name (e.g. bridge)
• variant (int): numeric variant from folder suffix (bridge_3 -> 3, base folder -> 1)
• scene_variant (string): source folder name
• frame_id (int): numeric frame id from filename
• image (struct): embedded image bytes + relative path
• image_relpath (string): relative source image path
• params_relpath (string): relative source JSON path
• raw_params_json (string): full original JSON text
• camera_json (string): camera section
• actors_json (string): actors section
• source_json (string): source section
• actor_labels (list[string]): unique labels found in actors
• has_label_mapping (bool): whether source.label_mapping exists
• label_mapping_json (string): full mapping JSON
• label_mapping_keys (list[string]): mapping keys
• label_mapping_values (list[string]): mapping values
• original_params_name (string): source.original_params when present
• upload_batch_utc (string): UTC timestamp of upload run

Loading

from datasets import load_dataset, Image

ds = load_dataset("turhancan97/SpaRRTa", split="train")
# Convert struct column to datasets Image feature if needed:
ds = ds.cast_column("image", Image())

Download to Local Machine

Download the dataset repo files (parquet shards, README, manifest):

huggingface-cli download turhancan97/SpaRRTa --repo-type dataset --local-dir ./hf_SpaRRTa

Rebuild Original Folder Structure

The script below reconstructs: <output>/<scene_variant>/img_XXXX.jpg and params_XXXX.json

from pathlib import Path
from datasets import load_dataset, Image

repo_id = "turhancan97/SpaRRTa"
output_root = Path("position_between_objects")
output_root.mkdir(parents=True, exist_ok=True)

ds = load_dataset(repo_id, split="train")
ds = ds.cast_column("image", Image(decode=False))

for row in ds:
    mid = output_root / row["scene_variant"]
    mid.mkdir(parents=True, exist_ok=True)

    image_name = Path(row["image_relpath"]).name
    params_name = Path(row["params_relpath"]).name

    image_bytes = row["image"]["bytes"]
    if image_bytes is None:
        raise RuntimeError(f"Missing embedded image bytes for {row['sample_id']}")

    (mid / image_name).write_bytes(image_bytes)
    (mid / params_name).write_text(row["raw_params_json"], encoding="utf-8")

Citation

If you find this research useful, please consider citing:

@misc{kargin2026sparrta,
  title={SpaRRTa: A Synthetic Benchmark for Evaluating Spatial Intelligence in Visual Foundation Models},
  author={Turhan Can Kargin and Wojciech Jasiński and Adam Pardyl and Bartosz Zieliński and Marcin Przewięźlikowski},
  year={2026},
  eprint={2601.11729},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2601.11729}
}