euler-ds-info

Dataset-level statistics helpers for Euler-view.

Current mode:

• estimate-mor - estimate a lidar-aware MOR profile from RGB, sparse depth, intrinsics, and camera extrinsics loaded through euler-loading.

Installation

pip install euler-ds-info

Usage

python -m euler_ds_info --input-json - <<'JSON'
{
  "mode": "estimate-mor",
  "modalities": {
    "rgb": "/data/dataset/train#rgb",
    "sparse_depth": "/data/dataset/train#sparse_depth",
    "intrinsics": "/data/dataset/train#intrinsics",
    "camera_extrinsics": "/data/dataset/train#camera_extrinsics"
  }
}
JSON

The command prints a single JSON document with per-file MOR profiles and aggregate summary statistics.

The output includes a top-level glossary section. Metric definitions are shared between per_file_info and aggregate reducer outputs, and each metric entry includes display labels, units, value ranges, interpretation hints, and caveats for downstream visualization.

To generate a sample artifact for inspection, run:

python -m euler_ds_info.dev_artifacts

This writes a realistic sample JSON document to ./.outputs/euler-ds-info.sample.json by calling the real MOR pipeline against a mocked in-memory euler-loading dataset.

Performance

• The CLI will use SLURM_CPUS_PER_TASK automatically when present.
• You can override the worker count with --workers N or EULER_DS_INFO_WORKERS=N.
• The loader still runs sequentially; the speedup comes from parallel per-sample profiling.