BubbleID

BubbleID is a research model package for analyzing liquid-vapor interfaces in high-speed pool-boiling image sequences. The canonical software repository is cldunlap73/BubbleID; it provides installation, runtime, and tutorial guidance. This Hugging Face repository hosts the two pretrained checkpoint files released with the citable BubbleID data record.

Included checkpoints

File Purpose SHA-256
seg_model.pth Pretrained instance-segmentation checkpoint FAA9DDAAE71C186AB749ED33BA18EDE07D0B24CD75D23B8F9FC880CB3467FF42
class_model.pth Pretrained classification checkpoint DBAEF1211F87D9C41340301577CB19D076F1513DBFE37B8918251E7831AF0AA0

Intended use

The release is intended for research use with pool-boiling image sequences. BubbleID combines segmentation, tracking, and classification workflows to support image-derived bubble/interface analysis. Use the versioned GitHub repository for dependencies, model invocation, and tutorial steps.

The source release documents test sequences sampled at 3000 fps (steady state) and 150 fps (transient). Frame-rate-dependent outputs and model behavior should be interpreted using the source documentation and experimental metadata.

Data and provenance

The model files originate from the Dryad dataset Data from: BubbleID: A deep learning framework for bubble interface dynamics analysis, published May 23, 2025:

The full 946.59 MB archival package, including annotations and test sequences, remains on Dryad to preserve its citable version. This Model repository intentionally does not duplicate those data.

Validation status and limitations

  • The checkpoint SHA-256 values above were verified against a local copy of the Dryad release.
  • Checkpoint loading and end-to-end inference were not rerun for this Hugging Face release: the release-preparation environment did not have PyTorch installed.
  • The models were released for pool-boiling image data; transfer to other facilities, optics, fluids, surfaces, frame rates, or boiling regimes requires independent evaluation and, where needed, fine-tuning.
  • This repository is not an inference endpoint and does not claim Hugging Face Inference Provider support.

Attribution and licensing

Please cite the Dryad dataset and associated BubbleID publication when using these files. The originating GitHub repository lists code licenses; that alone does not establish a separate license for model weights or data derivatives. Accordingly, this repository uses the other license designation and links directly to the source record. Contact the listed source-record authors for reuse questions beyond the documented release.

Citation

Dunlap, Christy; Li, Changgen; Pandey, Hari; Le, Ngan; Hu, Han (2025). Data from: BubbleID: A deep learning framework for bubble interface dynamics analysis [Dataset]. Dryad. https://doi.org/10.5061/dryad.ksn02v7gx

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support