Skip to content

ppg-hybrid-16x

Overview

  • Modality: PPG
  • Structure: codec
  • Compression ratio: 16×
  • Sample rate: 64 Hz
  • Config: — (operating point is fully declared in the registry; no training config)
  • Run name: ppg_hybrid_64hz_16x_golden
  • HuggingFace: Ambiq/compressionkit-ppg-hybrid-16x-v1.0

Dataset & License

  • Dataset: Open unified PPG v1 (dataset_id: ppg-unified-strict-sanitize-v1)
  • License: Open (BIDMC, BUT PPG, PPG-DaLiA, WESAD — mixed open licenses, no restricted-access dependency)
  • Notes: Sources: BIDMC, BUT PPG, PPG-DaLiA, and WESAD. Published v1 PPG goldens are MESA-free. Build the cache with scripts/build_ppg_cache.py.

The lifecycle runner pre-flights dataset availability before training (see #26 and the dataset contract).

Reproduction

# Single command, end-to-end.
uv run compressionkit golden run ppg-hybrid-16x

# Publish the deploy package to HuggingFace (requires HF_TOKEN).
uv run compressionkit golden run ppg-hybrid-16x --publish

Results land under results/ppg_hybrid_64hz_16x_golden/; deploy artifacts under results/ppg_hybrid_64hz_16x_golden/deploy/.

Evaluation Metrics

See the modality model zoo for the full metrics table:

Each run writes quality_scorecard.json and summary.json under its results/<run>/.

Deploy Artifacts

Every successful run produces the canonical edge deploy package:

  • spiht_config.json / spiht_app_config.h — codec parameters (language-neutral + C header).
  • c_sources/spiht.[ch] — portable C99 SPIHT reference.
  • sample_stimulus.npz / reference_vectors.npz — license-safe test frames and known-good encode/decode vectors.
  • model_card.json, deploy_manifest.json — metadata.
  • denoiser_gain_model.tflite / .h — INT8 wavelet-gain denoiser (embeddable, LiteRT).
  • denoiser_gain_model.keras, hybrid_manifest.json — float32 Python reference denoiser and pipeline stage order.

Customization Notes

  • This operating point is declared directly in compressionkit/experiments/registry.py (no training YAML) — add a new registry entry to explore a neighbouring operating point.
  • To resume publishing without retraining, pass --skip-train to compressionkit golden run ppg-hybrid-16x --publish.