Fast2Peak: Datasets and Supplementary Videos

This repository hosts data and supplementary videos for Fast2Peak (F2P). The six benchmark HDF5 files were migrated without numerical modification from the former F2P-Benchmark repository into the layout below. The superseded repositories have been removed; this repository is the current location for these data. Migration provenance and file checksums are recorded in data_manifest.json.

datasets/
  train/
    train.h5
    val.h5
  test/
    Simu.h5
    2DMs.h5
    Hela.h5
    SSEs.h5
  experiments/
real-time denoising video/
data_manifest.json
README.md

Availability

• Training, validation and the four benchmark test files are available.
• datasets/experiments/ is reserved for additional experimental/application data; upload is pending.
• real-time denoising video/ is reserved for real-time imaging/denoising videos; upload is pending.
• F2P and ResUNet weights with their original configurations, plus the historical SMAE benchmark checkpoint with reconstructed evaluation settings, are available in the separate model repository.

datasets/test/Simu.h5 is the original uploaded Simu benchmark file. It is not the separately constructed instrument-noise SFT test set. No re-noising, preprocessing or split changes were performed during this migration.

Download

While this repository is private, authentication with access to the repository is required. From the Fast2Peak project root, download only the asset directories to avoid overwriting the project's README:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="moria5161/fast2peak",
    repo_type="dataset",
    local_dir=".",
    allow_patterns=["datasets/**", "real-time denoising video/**"],
)

Original paths, the source revision, file sizes and SHA-256 checksums are recorded in data_manifest.json.

Code and Citation

Code and usage instructions: moria5161/fast2peak on GitHub. Manuscript citation details will be added when available.

Terms

Dataset and video licensing information will be finalized before public release. The GitHub code's Apache-2.0 license does not automatically apply to these assets.