Attention:The NSF Public Access Repository (PAR) system and access will be unavailable from 11:00 PM ET on Thursday, November 12th until 12:00 AM ET on Friday, November 13th due to maintenance. We apologize for the inconvenience.


Title: BoilingBench-CV Dataset
Abstract: Overview    BoilingBench-CV is a benchmark dataset package for evaluating computer-vision methods for two-dimensional bubble instance segmentation and per-frame bubble morphometry. It is designed to measure within-domain performance and generalization across water and HFE-7100 imaging domains. This package does not yet establish a tracking or bubble-dynamics benchmark. Per-image contours do not encode reviewed temporal bubble identities, coalescence/breakup events, or departure times. Contents - `raw/`: immutable copies of source annotated-image subsets and their original annotation formats. - `derived/`: canonical COCO-style annotations and manifests generated from the raw data. - `splits/`: immutable grouped benchmark partitions. - `metadata/`: source, rights, annotation, and quality-assurance records. - `docs/`: data-card and release notes. The source video collections are deliberately not duplicated here. The benchmark uses annotated still images; original videos remain in their source archives. Data domains The initial pool-boiling component includes FCu-H2O, PCu-H2O, PSi-HFE, and SSi-HFE annotated subsets. They form a valuable but compound domain shift: fluid, surface, image geometry, optical arrangement, and facility may differ. Water-versus-HFE results must not be interpreted as a fluid-only causal effect. Use rules 1. Preserve `raw/` unchanged. 2. Use only the supplied grouped split files for reported benchmark results. 3. Do not split frames from the same `(regime, source video/power condition)` across partitions. 4. Report performance by regime, not only a pooled score. 5. Treat annotations flagged in `metadata/annotation_qa.md` according to the    benchmark evaluator's documented ignore policy. Attribution When using the package, cite the BoilingBench-CV dataset version and the source dataset records listed in `metadata/data_sources.csv`, together with the assigned Zenodo DOI: [10.5281/zenodo.22264378](https://doi.org/10.5281/zenodo.22264378).  more » « less
Award ID(s):
2323022
PAR ID:
10712566
Author(s) / Creator(s):
; ; ; ; ; ;
Publisher / Repository:
Zenodo
Date Published:
Edition / Version:
v0.1.0
Subject(s) / Keyword(s):
Right(s):
Creative Commons Attribution 4.0 International
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract: BoilingBench-Multimodal Lite is a versioned, machine-learning-oriented release from the NED³ Laboratory at the University of Arkansas. It contains seven related benchmark tracks covering pool boiling, bubble-image annotation, acoustic sensing, and infrared immersion cooling. BoilingBench-1 through BoilingBench-4 contain multimodal boiling experiments; BoilingBench-5 contains human-annotated boiling images; BoilingBench-6 is a hydrophone-focused pool-boiling dataset; and BoilingBench-7 is an infrared immersion-cooling dataset. This Lite profile omits original Phantom `.cine` recordings to reduce distribution size. It retains the released raw files, processed analysis products, annotations, manifests, plots, and machine-readable exports included in the directory. The complete archive and any omission details are documented in the accompanying README and release manifest. Raw acquisition files are measurements or source observations. Heat flux, reconstructed surface temperature, synchronization records, spectra, spectrograms, event times, hysteresis metrics, and other analysis products are derived data unless explicitly identified otherwise. BoilingLab-generated heat-flux products require users to inspect thermocouple locations, geometry, calibration, material assumptions, time alignment, sign convention, and quality fields before treating them as labels. BoilingBench-3 and BoilingBench-4 use legacy MISTRAS USB AE acquisition and do not contain continuous AE waveform exports. BoilingBench-3 contains microphone data. BoilingBench-1 and BoilingBench-2 contain continuous AE waveform products where documented. Users must split examples by independent experiment/run or physical condition rather than randomly splitting adjacent temporal windows. The dataset is intended for reproducible research in heat-transfer prediction, multimodal sensing, boiling-regime and transition analysis, bubble-image analysis, acoustic heat-flux estimation, sequence regression, and cross-domain generalization. It is not a clinical, personal, or safety-certified dataset. 
    more » « less
  2. Abstract: Overview   This dataset supports the manuscript *BubbleID-Flow: Machine-Vision Quantification of Vapor Area Fraction in Subcooled Flow Boiling*. It contains high-speed optical images, manually annotated images, thermal workbooks, acoustic-emission (AE) records, and the fine-tuned Mask R-CNN checkpoint used in the reported analysis. The experiments used FC-72 in a rectangular flow passage that was 2.5 mm wide and 5.0 mm high. The bottom wall was electrically heated over 114.6 mm. High-speed images were acquired with a FASTCAM SA-Z at 1024 x 1024 pixels and 3000 frames/s. The manuscript reports a two-dimensional, near-wall vapor area fraction measured from the image-plane union of predicted vapor masks. It is not a volumetric vapor fraction or void fraction. Contents ```text AnnotatedImages/   130 paired BMP images and LabelMe JSON annotation files. ModelWeights/   model_final.pth                 Fine-tuned one-class Mask R-CNN checkpoint. MultimodalData/   Images/                         Raw high-speed images and camera metadata.   FlowLoopThermalData/            Thermal workbooks and derived thermal plots.   AcousticDatta/                  AE event, continuous, waveform, and plot files. Videos/   BubbleID-Flow_45V_raw_vs_segmented.mp4                                   Derived 16-second presentation video.   BubbleID-Flow_45V_raw_vs_segmented.json   Video frame-selection and inference manifest.   BubbleID-Flow_45V_raw_top_segmented_bottom.mp4                                   Derived 16-second presentation video.   BubbleID-Flow_45V_raw_top_segmented_bottom.json                                   Video frame-selection and inference manifest. ``` `AnnotatedImages` contains 130 valid LabelMe JSON files paired with 130 BMP images. The annotation labels are `bubble` and `bubble_cluster`; the BubbleID-Flow model merges both labels into one vapor-region class for union-mask vapor-area-fraction analysis. `MultimodalData/Images` contains 3,977 BMP images and 37 `.cihx` camera metadata files. `MultimodalData/FlowLoopThermalData` contains three Excel workbooks per case: `.xlsx`, `_IN_with_units.xlsx`, and `_u_with_units.xlsx`. The PNG files under `Plot Images` and the AE `Plot_Hit` and `Plot_Time` folders are derived visualizations, not primary measurement inputs. `Videos/BubbleID-Flow_45V_raw_vs_segmented.mp4` is a derived presentation output. It shows 20 uniformly sampled frames from the 45 V state of each case at 5 frames/s (16 seconds total), with a raw near-wall view and the corresponding Mask R-CNN union-mask overlay. Its JSON sidecar records the selected filenames, checkpoint checksum, ROI, score threshold, and detection cap. Case Map | Canonical case ID | Nominal mass flow | Nominal inlet subcooling | Optical image folder | Thermal folder | AE folder | | --- | ---: | ---: | --- | --- | --- | | `5gs_22C` | 5 g/s | 22 degC | `MultimodalData/Images/5gs_22C` | `MultimodalData/FlowLoopThermalData/5gs_22C` | `MultimodalData/AcousticDatta/5gs_22C` | | `10gs_22C` | 10 g/s | 22 degC | `MultimodalData/Images/10gs_22C` | `MultimodalData/FlowLoopThermalData/10gs_22C` | `MultimodalData/AcousticDatta/10gs_22C` | | `15gs_20C` | 15 g/s | 20 degC | `MultimodalData/Images/15gs_20C` | `MultimodalData/FlowLoopThermalData/15gs_20C` | `MultimodalData/AcousticDatta/15gs_22C` (archive alias) | | `25gs_20C` | 25 g/s | 20 degC | `MultimodalData/Images/25gs_20C` | `MultimodalData/FlowLoopThermalData/25gs_20C` | `MultimodalData/AcousticDatta/25gs_22C` (archive alias) | Archive naming notes - `AcousticDatta` is the current directory name in this archive. - The AE directories `15gs_22C` and `25gs_22C` are aliases only. Their contained HIT, TIME, DTA, and WFS filenames identify the canonical `15gs_20C` and `25gs_20C` cases. Use the canonical IDs in analyses and citations. - Optical folder labels such as `25_ONSET`, `55 CHF`, and `59CHF` are source acquisition labels. They are not independently validated onset or critical-heat-flux labels. Modality Details Optical images The image folders are organized as `/`. The study uses a fixed near-wall region of interest (ROI) beginning at pixel coordinate `(0, 485)` with size `1024 x 70` pixels. Spatial calibration and registration between this ROI, the heated surface, and the seven temperature locations were not available. Therefore, this archive does not support calibrated bubble-size measurements, local optical-thermal registration, or volumetric void-fraction calculation. Thermal data The thermal workbooks contain time, mass flow, voltage, electrical power, inlet subcooling, inlet/outlet temperature, pressure, and seven test-section surface-temperature streams. The manuscript derives heat flux from electrical power and the documented heated area and derives mass flux from the measured flow area. Reported standard deviations characterize temporal variability within selected windows; they are not instrument uncertainty estimates. For each canonical case, the BubbleID-Flow manuscript pipeline uses both: - `.xlsx` for thermal variables; and - `_IN_with_units.xlsx` to obtain voltage and define/match state windows. Acoustic-emission data Each AE case directory contains: - `HIT_.TXT`: event-level hit data; - `TIME_.TXT`: continuous time-domain summary data; - `Raw Data/.wfs`: raw waveform file; - `Raw Data/.DTA`: acquisition/export file; - `notes.txt`: available operator notes; and - `Plot_Hit` and `Plot_Time`: derived diagnostic figures. The optical, thermal, and AE systems do **not** have a verified common hardware trigger. The manuscript's supplementary 10 g/s AE screen uses a 25.395 s archive-start timestamp offset for descriptive state comparison only. These data do not establish optical-acoustic synchronization, causal coupling, source localization, flow-regime classification, or calibrated AE energy. Annotations and model checkpoint The model was fine-tuned on 130 manually annotated images. The published reconstruction evaluation uses a deterministic same-sequence split: every fifth sorted image is assigned to holdout, yielding 104 training images and 26 holdout images. This is not an experiment-held-out generalization test; the source-sequence proximity between training and holdout images must be considered when interpreting performance. The included checkpoint is: ```text ModelWeights/model_final.pth SHA-256: 6f60969ce876f57a78b53fc61895c99b92cb2e1b6f0e0b30f01fa301236042cd ``` 
    more » « less
  3. Abstract: This paper presents BubbleID, a sophisticated deep learning architecture designed to comprehensively identify both static and dynamic attributes of bubbles within sequences of boiling images. By amalgamating segmentation powered by Mask R-CNN with SORT-based tracking techniques, the framework is capable of analyzing each bubble's location, dimensions, interface shape, and velocity over its lifetime and capturing dynamic events such as bubble departure. BubbleID is trained and tested on boiling images across diverse heater surfaces and operational settings. This paper also offers a comparative analysis of bubble interface dynamics prior to and post-critical heat flux conditions. TechnicalInfo: # Data from: BubbleID: A deep learning framework for bubble interface dynamics analysis Dataset DOI: [10.5061/dryad.ksn02v7gx](10.5061/dryad.ksn02v7gx) ## Description of the data and file structure This dataset includes raw bubble images, annotated images, and saved model weights for [BubbleID](https://github.com/cldunlap73/BubbleID), an open-source package for bubble identification, tracking, and analysis, and liquid-vapor interface dynamics analysis. The bubble images are collected using a high-speed camera (Phantom VEO 710L) during pool boiling tests performed using the facility described in [Pandey et al. 2024.](https://www.sciencedirect.com/science/article/pii/S1359431123018367) The bubble images are manually annotated for the liquid-vapor interfaces using the [labelme](https://anaconda.org/conda-forge/labelme) package.  ### Files and variables #### File: ned3-001\_BubbleID.zip **Description:** The zip file includes four folders, i.e., AnnotatedData, Models, TestData, and Tutorials.  The *AnnotatedData* folder includes two subfolders and a readme file. This dataset contains pool boiling image data and annotations made in labelme. * Classes-1: Annotations for liquid-vapor interfaces, including the training set, testing set, and a set from one pool boiling experiment, which does not appear in the other folders (Unseen-Test). - Classes-2: Annotations for bubbles distinguishing between attached and detached bubbles. The *Models* folder includes PyTorch model checkpoint file (.pth) saving the weights of preitrained models, including * seg_model.pth: pre-trained instance segmentation model * class_model.pth: pre-trained classification model The *TestData* includes two subfolders:  * SteadyStateTest for image sequences sampled at 3000 fps.  * TransientSmallTst for image sequences sampled at 150 fps.  The *tutorial* folder includes a Python Notebook script illustrating the steps for using the pretrained model and the BubbleID class for saving and displaying data extracted from bubble images/ videos.  ## Code/software * The images (.jpg) and annotations (.json) can be viewed using the [labelme](https://anaconda.org/conda-forge/labelme "https://anaconda.org/conda-forge/labelme") package.  Yes (and it should be a simple as installing LabelMe and pointing to the directory of the images and JSON files. * The tutorial (.ipynb) can be opened using any Python Notebook packages, e.g., Jupyter Notebook, Google Colab. It can be viewed in Google Colab, but I don't think there is a way (at least not a simple one) to run it in Colab. It can be viewed in Jupyter Notebook and run there if installed correctly * The [PyTorch](https://pytorch.org/ "https://pytorch.org/") package can import the checkpoint file (.pth) in a Python script. This can be done by specifying the path to the weights using the BubbleID class, e.g., *test120=BubbleID.DataAnalysis(imagesfolder,videopath,savefolder,extension,modelweights,"cpu")* 
    more » « less
  4. DescriptionThis record contains filtered single cell RNA sequencing data from Dictyostelium discoideum used in the manuscript: “Early cyclic adenosine monophosphate signaling orchestrates single cell synchronicity throughout Dictyostelium development” by Katoh Kurasawa, Trnovec et al. FilesThree AnnData objects in .h5ad format (compatible with Python and Scanpy) are provided: AX4.h5ad acaA.h5ad acaA_pkaC.h5ad Each file includes the filtered expression matrix and associated cell and gene metadata used for downstream analyses in the manuscript, including clustering, annotation, trajectory analyses, and synchronicity quantification. Relationship to raw dataRaw sequencing data and primary submission files are available through the Gene Expression Omnibus repository under accession [GSE305468]. Reproducible analysis codeThe analysis pipeline corresponding to this dataset is archived as Zenodo software record https://doi.org/10.5281/zenodo.18481808 and is also available on GitHub at https://github.com/lenatr99/scRNA_dicty. NotesThese .h5ad files represent processed and filtered data intended for analysis and figure reproduction. For full raw data and submission metadata, please use the Gene Expression Omnibus accession listed above. 
    more » « less
  5. Abstract: African elephants are a species of conservation concern with complex vocal behavior. Annotated acoustic recordings of elephants can be used both for studying vocal behavior, including call rates and types, and developing acoustic detection and classification models. We present an annotated dataset of African elephant vocalizations from an affordable acoustic recorder (AudioMoth) integrated into a GPS collar deployed onto a wild female African elephant in Kenya. A total of 62 hours of animal-borne acoustic data were recorded and 1,123 vocalizations were identified including 706 high-quality rumbles that were further used in acoustic feature analysis. The average group call rate was 18.1 calls/hr (range: 0 - 104), with a change point analysis identifying three periods of declining call rate after the collaring event. Our findings offer insight into the vocal behavior of African elephants, and the data provided can be used in the development of acoustic detection models. TechnicalInfo: # Data from: Animal-borne acoustic recordings of African elephant vocalizations from Samburu National Reserve, Kenya This dataset contains audio recordings collected using an AudioMoth recorder deployed onto a GPS collar on an African elephant named Arden in Samburu, Kenya, in June 2025. ## Folder structure **audio.zip** (3.73 GB) - zipped folder containing 62 one-hour .wav files recorded at a 16 kHz sample rate. The start time of each file (local time, East Africa Time) is encoded in the filename, e.g. `Arden_20250606_170125.WAV`. The beginning of the filename indicates the name of the individual (Arden). **clips.zip** (120.8 MB) - zipped folder containing 1,123 .wav clips at the original sample rate (16 kHz), each containing one manually identified rumble extracted from the raw audio with a 0.2 s buffer added before and after the call. Each clip filename includes the source recording, the selection number of the call within that recording, the call type, and a quality indicator. **labels.zip** (52.8 KB) - contains 62 .txt annotation Raven files, one per raw audio file, with matching filenames. Three recordings had no vocalizations identified during manual review; empty annotation files are included for these so that the file counts between the raw audio and label folders match. **model.zip** (403.75 KB) - contains the automated rumble detection model output, run separately on 8 kHz audio and 4 kHz audio. Each of the two `4kHz` and `8kHz` subfolders contains: * a combined **annotation** file with one row per detection across all 62 recordings * a **summary** file with one row per recording, giving the file duration and total detection counts * a **per_file_annotations** subfolder, containing the per-recording split of the combined annotation file (one .txt file per source recording, matching the manual label filenames). These annotations can be directly opened and viewed in Raven. ## Manual annotation file format Each manual label file is a tab-delimited text file with one row per annotated call and the following columns: * **Selection** - ID number assigned to the call. Numbers may skip if a selection was removed during review. * **View** - spectrogram view label, constant across all files (Spectrogram 1). * **Channel** - recording channel, always 1 since all recordings are mono. * **Begin Time (s) / End Time (s)** - start and end time of the call, in seconds from the start of the recording. * **Low Freq (Hz) / High Freq (Hz)** - lower and upper frequency bounds of the annotation box. * **call_type** - RUM (rumble), ROAR, or ROAR-RUM (a call containing both). * **quality** - overall quality rating of the call, on a 0-4 scale (see Table 2 in the associated manuscript for the full description). * **overlap** - whether the call overlapped in time with another vocalization (Y/N). * **earflap** - presence of acoustic interference from ear flapping, coded 0-2 (see Table 2 in the associated manuscript for the full description). All annotations were made manually in Raven Pro. ## Model output format ### Annotation files `model/4kHz/audiomoth_downsampled_4kHz_annotation.txt` (1136 detections, 4 kHz audio) and `model/8kHz/audiomoth_downsampled_8kHz_annotation.txt` (1,395 detections, 8 kHz audio) contain the automated call detections from the trained model across all 62 recordings combined. These files use the same tab-delimited Raven selection table format as the manual labels, with additional columns generated by the detection pipeline: * **Selection** - ID number assigned to the detection. * **View** - spectrogram view label, constant across all rows. * **Channel** - recording channel; always 1 (mono). * **Event** - class label assigned by the model (Rumble for all detections in this dataset). * **Begin Time (s) / End Time (s)** - start and end time of the detection, in seconds from the start of the full concatenated recording stream (not the individual file). * **Low Freq (Hz) / High Freq (Hz)** - lower and upper frequency bounds of the detection. * **Begin Path** - full file path to the source audio file at the time the model was run. * **File Offset (s)** - start time of the detection, in seconds from the start of the individual audio file named in Begin File. * **Begin File** - filename of the audio file containing the start of the detection. * **Site** - recording site/device name. For this dataset, this is the name of the collared elephant (Arden). * **Begin Hour** - local hour of day (0-23) when the detection begins. * **File Start Date / Begin Date** - calendar date (local time) of the file start and of the detection. * **Score** - model confidence score for the detection. * **Count, Measurable, Harmonics, Ambiguous, Notes, Analyst, Rand, Deployment, Sound Problems, Call Criteria, Disk** - fields carried over from the Raven annotation template; left blank in the automated output and not used for the model results. ### Summary files `model/4kHz/audiomoth_downsampled_4kHz_summary.txt` and `model/8kHz/audiomoth_downsampled_8kHz_summary.txt` each contain one row per source recording, with the following tab-delimited columns: * **Begin Path** - full file path to the source audio file at the time the model was run. * **File Duration (s)** - length of the recording in seconds. * **Rumble Count** - number of rumble detections in that recording. * **Gunshot Count** - number of gunshot detections in that recording (0 throughout this dataset; the model was not applied to a gunshot-annotated training set for this deployment). ### Per-file annotations `model/4kHz/per_file_annotations/` and `model/8kHz/per_file_annotations/` each contain one .txt file per source recording (e.g. `Arden_20250606_130550_annotations.txt`), giving the subset of the combined annotation file corresponding to that recording, in the same column format described above. Unlike the manual `labels` folder, recordings with zero model detections do not have a corresponding file in these subfolders (59 files for 4 kHz; 61 files for 8 kHz). 
    more » « less