Find research datasets worth reusing
Search datasets from major research repositories and use ShareScore to quickly assess how well each record supports discovery, access, and reuse.
42
datasets available to search
ShareScore release 0.9.0
Dataset results
42 results for “DCASE”
DCASE 2024 Task 9: Language-Queried Audio Source Separation | Evaluation Set
<p>This is the <strong>evaluation set for Task 9, Language-Queried Audio Source Separation (LASS), in DCASE 2024 Challenge</strong>. </p> <p>This evaluation set is meant to be used for Task 9 at the scientific challenge DCASE 2024. This split is not meant to be used for training LASS methods. This split is meant to be used for evaluating LASS methods in the final testing & ranking stage. All audio clips are sourced from Freesound, uploaded between April and October 2023. Each audio file has been segmented into 10-second clips and converted to mono 16 kHz.</p> <p>This evaluation set consists of<strong> evaluation set (synth)</strong> and an <strong>evaluation set (real)</strong>. </p> <p><strong>== Evaluation set (synth) ==</strong></p> <p>This evaluation set is created using 1,000 audio clips. Each clip is annotated with three captions describing the content of the clip. We created 3,000 synthetic mixtures with signal-to-noise ratios (SNR) ranging from -15 to 15 dB. Each synthetic mixture includes one natural language query and its corresponding target source. We used annotated tag information to ensure that the two audio clips used in each mix do not share overlapping sound source classes. The original audio files used to create these mixtures are not released. The mixtures and language queries are available for evaluation.</p> <p>The audio files in the archives:</p> <ul> <li>lass_evaluation_synth.zip</li> </ul> <p>and the associated metadata (including audio filename and text queries) in the CSV file:</p> <ul> <li>lass_synthetic_evaluation.csv</li> </ul> <p><strong>== Evaluation set (real) ==</strong></p> <p>This evaluation set consists of 100 audio clips. Each audio clip contains at least two overlapping sound sources. For each audio clip, we manually annotated their component sources using text descriptions, so that each clip can be used as a 'mixture' from which to extract one or more of the component sources based on a text query. Each audio clip in evaluation (real) was labeled with two such text queries.</p> <p>The audio files in the archives:</p> <ul> <li>lass_evaluation_real.zip</li> </ul> <p>and the associated metadata (including audio filename and text queries) in the CSV file:</p> <ul> <li>lass_real_evaluation.csv</li> </ul>
DCASE 2023 Task5: Few-shot Bioacoustic Event Detection: Evaluation set
<p><strong>General Description</strong></p> <p>The evaluation set for task 5 of DCASE 2023 "Few-shot Bioacoustic Event Detection" consists of 8 subsets of data representing different acoustic sources, in total there are 66 audio files. </p> <p>The first 5 annotations are provided for each file, with events marked as positive (POS) for the class of interest. </p> <p>This dataset is to be used for evaluation purposes during the task.</p> <p> </p> <p><strong>Folder structure</strong></p> <p>Audiofiles and annotation files are organized across 3 different zip files:</p> <p><em>Eval_1.zip</em></p> <p> |___CHE23/</p> <p> |____*.wav</p> <p> |____*.csv</p> <p> |___CW/</p> <p> |____*.wav</p> <p> |____*.csv</p> <p><em>Eval_2.zip</em></p> <p> |___MGE/</p> <p> |____*.wav</p> <p> |____*.csv</p> <p> |___MS/</p> <p> |____*.wav</p> <p> |____*.csv</p> <p> |___QU/</p> <p> |____*.wav</p> <p> |____*.csv</p> <p> <em>Eval_3.zip</em></p> <p> |___DC/</p> <p> |____*.wav</p> <p> |____*.csv</p> <p> |___CT/</p> <p> |____*.wav</p> <p> |____*.csv</p> <p> |___CHE/</p> <p> |____*.wav</p> <p> |____*.csv</p> <p><em>Annotations_only.zip</em> contains the *.csv files organised in the same subfolders.</p> <p>*The subfolders denote different recording sources and there may or may not be overlap between classes of interest from different wav files.</p> <p>*Note that there can be different target classes within the same subfolder</p> <p> </p> <p><strong>Annotation structure</strong></p> <p>Each line of the annotation csv represents an event in the audio file. The column descriptions are as follows:<br> [ Audiofilename, Starttime, Endtime, Q ]</p> <p><strong>Development Set</strong></p> <p>The development set for the same task can be found at: <a href="http://doi.org/10.5281/zenodo.4543504">https://doi.org/10.5281/zenodo.6012309</a>. </p> <p><strong>Open Access</strong></p> <p>This dataset is available under a Creative Commons Attribution 4.0 International (CC BY 4.0) license.<br> </p> <p><strong>Contact info</strong></p> <p>Please send any feedback or questions to:<br> Ines Nolasco: i.dealmeidanolasco@qmul.ac.uk</p> <p>or join us on slack: <a href="https://join.slack.com/t/dcase/shared_invite/zt-12zfa5kw0-dD41gVaPU3EZTCAw1mHTCA">task-fewshot-bio-sed</a></p>
Evaluation set DCASE 2023 task 4 (for submissions)
<p>This repo contains the dataset to download to submit results and be evaluated in task 4 of DCASE 2023. It also contains the ground-truth for the public and synthetic evaluation dataset, together with the mapping file between the anonymized (official eval) file names and the files name as presented in the annotations.</p> <p>Please, check the submission package in order to follow the instruction to have a submission.</p>
Submissions DCASE 2023 Task4a
<p>Predictions and technical reports of the systems submitted to <a href="https://dcase.community/challenge2023/task-sound-event-detection-with-weak-labels-and-synthetic-soundscapes">DCASE 2023 Task4a</a> including file name mapping and ground truth for public youtube evaluation set. Challenge results can be found on <a href="http://dcase.community/challenge2023/task-sound-event-detection-with-weak-labels-and-synthetic-soundscapes-results">results page</a> and additional post-processing independent evaluations can be found in [1].</p> <p>[1] J. Ebbers, R. Haeb-Umbach, and R. Serizel, "Post-Processing Independent Evaluation of Sound Event Detection Systems", Detection and Classification of Acoustic Scenes and Events (DCASE) Workshop, 2023, arXiv: <a href="https://arxiv.org/abs/2306.15440">https://arxiv.org/abs/2306.15440</a></p>
DCASE 2020 Challenge Task 2 Development Dataset
<p><strong>Description</strong></p> <p>This dataset is the "development dataset" for the <strong>DCASE 2020 Challenge Task 2 "Unsupervised Detection of Anomalous Sounds for Machine Condition Monitoring" </strong><a href="http://dcase.community/challenge2020/task-unsupervised-detection-of-anomalous-sounds">[task description]</a>. </p> <p>The data comprises parts of <strong>ToyADMOS</strong> and the <strong>MIMII Dataset</strong> consisting of the normal/anomalous operating sounds of six types of toy/real machines. Each recording is a single-channel (proximately) 10-sec length audio that includes both a target machine's operating sound and environmental noise. The following six types of toy/real machines are used in this task:</p> <ul> <li>Toy-car (ToyADMOS)</li> <li>Toy-conveyor (ToyADMOS)</li> <li>Valve (MIMII Dataset)</li> <li>Pump (MIMII Dataset)</li> <li>Fan (MIMII Dataset)</li> <li>Slide rail (MIMII Dataset)</li> </ul> <p> </p> <p><strong>Recording procedure</strong></p> <p>The ToyADMOS consists of normal/anomalous operating sounds of miniature machines (toys) collected with four microphones, and the MIMII dataset consists of those of real-machines collected with eight microphones. Anomalous sounds in these datasets were collected by deliberately damaging target machines. For simplifying the task, we used only the first channel of multi-channel recordings; all recordings are regarded as single-channel recordings of a fixed microphone. The sampling rate of all signals has been downsampled to 16 kHz. From ToyADMOS, we used only IND-type data that contain the operating sounds of the entire operation (i.e., from start to stop) in a recording. We mixed a target machine sound with environmental noise, and only noisy recordings are provided as training/test data. For the details of the recording procedure, please refer to the papers of <a href="https://ieeexplore.ieee.org/document/8937164">ToyADMOS</a> and <a href="http://dcase.community/documents/workshop2019/proceedings/DCASE2019Workshop_Purohit_21.pdf">MIMII Dataset</a>.</p> <p> </p> <p><strong>Data</strong></p> <p>We first define two important terms in this task: Machine Type and Machine ID. Machine Type means the kind of machine, which in this task can be one of six: toy-car, toy-conveyor, valve, pump, fan, and slide rail. Machine ID is the identifier of each individual of the same type of machine, which in the training dataset can be of three or four. Each machine ID's dataset consists of (i) around 1,000 samples of normal sounds for training and (ii) 100-200 samples each of normal and anomalous sounds for the test. The given labels for each training/test sample are Machine Type, Machine ID, and condition (normal/anomaly). Machine Type information is given by directory name, and Machine ID and condition information are given by their respective file names. </p> <p> </p> <p><strong>Directory structure</strong></p> <p>When you unzip the downloaded files from Zenodo, you can see the following directory structure. As described in the previous section, Machine Type information is given by directory name, and Machine ID and condition information are given by file name, as:</p> <ul> </ul> <p>/dev_data</p> <ul> <li>/ToyCar <ul> <li>/train (Only normal data for all Machine IDs are included.) <ul> <li>/normal_id_01_00000000.wav</li> <li>...</li> <li>/normal_id_01_00000999.wav</li> <li>/normal_id_02_00000000.wav</li> <li>...</li> <li>/normal_id_04_00000999.wav</li> </ul> </li> <li>/test (Normal and anomaly data for all Machine IDs are included.) <ul> <li>/normal_id_01_00000000.wav</li> <li>...</li> <li>/normal_id_01_00000349.wav</li> <li>/anomaly_id_01_00000000.wav</li> <li>...</li> <li>/anomaly_id_01_00000263.wav</li> <li>/normal_id_02_00000000.wav</li> <li>...</li> <li>/anomaly_id_04_00000264.wav</li> </ul> </li> </ul> </li> <li>/ToyConveyor (The other Machine Types have the same directory structure as ToyCar.)</li> <li>/fan</li> <li>/pump</li> <li>/slider</li> <li>/valve</li> </ul> <p> </p> <p>The paths of audio files are:</p> <ul> <li>"/dev_data/<Machine_Type>/train/normal_id_<Machine_ID>_[0-9]+.wav"</li> <li>"/dev_data/<Machine_Type>/test/normal_id_<Machine_ID>_[0-9]+.wav"</li> <li>"/dev_data/<Machine_Type>/test/anomaly_id_<Machine_ID>_[0-9]+.wav"</li> </ul> <p>For example, the Machine Type and Machine ID of "/ToyCar/train/normal_id_01_00000000.wav" are "ToyCar" and "01", respectively, and its condition is normal. The Machine Type and Machine ID of "/fan/test/anomaly_id_00_00000000.wav" are "fan" and "00", respectively, and its condition is anomalous.</p> <p> </p> <p><strong>Baseline system</strong></p> <p>A simple baseline system is available on the Github repository <a href="https://github.com/y-kawagu/dcase2020_task2_baseline">[URL]</a>. The baseline system provides a simple entry-level approach that gives a reasonable performance in the dataset of Task 2. It is a good starting point, especially for entry-level researchers who want to get familiar with the anomalous-sound-detection task.</p> <p> </p> <p><strong>Conditions of use</strong></p> <p>This dataset was created jointly by <strong>NTT Corporation</strong> and <strong>Hitachi, Ltd.</strong> and is available under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license.</p> <p> </p> <p><strong>Publication</strong></p> <p>If you use this dataset, please cite <strong>all the following three papers</strong>:</p> <p>Yuma Koizumi, Shoichiro Saito, Noboru Harada, Hisashi Uematsu, and Keisuke Imoto, "ToyADMOS: A Dataset of Miniature-Machine Operating Sounds for Anomalous Sound Detection," in Proc of Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA), 2019. <a href="https://ieeexplore.ieee.org/document/8937164">[pdf]</a></p> <p>Harsh Purohit, Ryo Tanabe, Kenji Ichige, Takashi Endo, Yuki Nikaido, Kaori Suefusa, and Yohei Kawaguchi, “MIMII Dataset: Sound Dataset for Malfunctioning Industrial Machine Investigation and Inspection,” in Proc. 4th Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE), 2019. <a href="http://dcase.community/documents/workshop2019/proceedings/DCASE2019Workshop_Purohit_21.pdf">[pdf]</a></p> <p>Yuma Koizumi, Yohei Kawaguchi, Keisuke Imoto, Toshiki Nakamura, Yuki Nikaido, Ryo Tanabe, Harsh Purohit, Kaori Suefusa, Takashi Endo, Masahiro Yasuda, and Noboru Harada, "Description and Discussion on DCASE2020 Challenge Task2: Unsupervised Anomalous Sound Detection for Machine Condition Monitoring<em>,"</em> in Proc. 5th Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE), 2020. <a href="https://dcase.community/documents/workshop2020/proceedings/DCASE2020Workshop_Koizumi_3.pdf">[pdf]</a></p> <p><br> <strong>Feedback</strong></p> <p>If there is any problem, please contact us:</p> <ul> <li>Yuma Koizumi, <a href="mailto:koizumi.yuma@ieee.org">koizumi.yuma@ieee.org</a></li> <li>Yohei Kawaguchi, <a href="mailto:yohei.kawaguchi.xk@hitachi.com">yohei.kawaguchi.xk@hitachi.com</a></li> <li>Keisuke Imoto, <a href="mailto:keisuke.imoto@ieee.org">keisuke.imoto@ieee.org</a></li> </ul> <p> </p>
Evaluation set DCASE 2020 task 4 (for submissions)
<p>This repo contains the dataset to download to submit results and be evaluated in task 4 of DCASE 2020.</p> <p> </p> <p>Please, check the submission package in order to follow the instruction to have a submission.</p> <p>*Note: some files are 5 mins long, so if your system is not suitable for this, make sure you aggegate the results.*</p>
DCASE 2021 AV Image Frames
Open the record for dataset details and reuse information.
DCASE 2024 Task 9: Language-Queried Audio Source Separation | Pre-trained Weights for the Baseline System
<p><strong>== Descriptions ==</strong></p> <p>We trained the AudioSep [1] model using the <a href="../records/10887496">development set</a> (Clotho and augmented FSD50K datasets) for 200k steps with a batch size of 16 using one Nvidia A100 GPU (around 1 day). Model details can be found in the <a href="https://arxiv.org/abs/2308.05037">AudioSep paper</a>.</p> <p>Pre-trained weights for the baseline system:</p> <ul> <li>audiosep_16k,baseline,step=200000.ckpt</li> </ul> <p>Baseline codebase:</p> <ul> <li>GitHub: <a href="https://github.com/Audio-AGI/dcase2024_task9_baseline">https://github.com/Audio-AGI/dcase2024_task9_baseline</a></li> </ul> <p><strong>== Reference ==</strong></p> <p>[1] Liu X, Kong Q, Zhao Y, et al. Separate anything you describe. arXiv:2308.05037, 2023.</p> <p><strong>== Contact ==</strong></p> <p>Xubo Liu, xubo.liu@surrey.ac.uk</p>
DCASE 2024 Challenge Task 2 Development Dataset
<p><strong><Data files will be made accessible on April 1st, 2024.></strong></p> <p><strong>Description</strong></p> <p>This dataset is the "development dataset" for the DCASE 2024 Challenge Task 2.</p> <p>The data consists of the normal/anomalous operating sounds of seven types of real/toy machines. Each recording is a single-channel 10-second audio that includes both a machine's operating sound and environmental noise. The following seven types of real/toy machines are used in this task:</p> <ul> <li>ToyCar</li> <li>ToyTrain</li> <li>Fan</li> <li>Gearbox</li> <li>Bearing</li> <li>Slide rail</li> <li>Valve</li> </ul> <p><strong>Overview of the task</strong></p> <p><strong>Anomalous sound detection (ASD) is the task of identifying whether the sound emitted from a target machine is normal or anomalous. </strong>Automatic detection of mechanical failure is an essential technology in the fourth industrial revolution, which involves artificial-intelligence-based factory automation. Prompt detection of machine anomalies by observing sounds is useful for monitoring the condition of machines. </p> <p>This task is the follow-up from DCASE 2020 Task 2 to DCASE 2023 Task 2. The task this year is to develop an ASD system that meets the following five requirements.</p> <p>1. **Train a model using only normal sound** (unsupervised learning scenario) <br>Because anomalies rarely occur and are highly diverse in real-world factories, it can be difficult to collect exhaustive patterns of anomalous sounds. Therefore, the system must detect unknown types of anomalous sounds that are not provided in the training data. This is the same requirement as in the previous tasks.</p> <p>2. **Detect anomalies regardless of domain shifts** (domain generalization task) <br>In real-world cases, the operational states of a machine or the environmental noise can change to cause domain shifts. Domain-generalization techniques can be useful for handling domain shifts that occur frequently or are hard-to-notice. In this task, the system is required to use domain-generalization techniques for handling these domain shifts. This requirement is the same as in DCASE 2022 Task 2 and DCASE 2023 Task 2.</p> <p>3. **Train a model for a completely new machine type** <br>For a completely new machine type, hyperparameters of the trained model cannot be tuned. Therefore, the system should have the ability to train models without additional hyperparameter tuning. This requirement is the same as in DCASE 2023 Task 2.</p> <p>4. **Train a model using a limited number of machines from its machine type** <br>While sounds from multiple machines of the same machine type can be used to enhance the detection performance, it is often the case that only a limited number of machines are available for a machine type. In such a case, the system should be able to train models using a few machines from a machine type. This requirement is the same as in DCASE 2023 Task 2.</p> <p>5 . **Train a model both with or without attribute information**<br>While additional attribute information can help enhance the detection performance, we cannot always obtain such information. Therefore, the system must work well both when attribute information is available and when it is not.</p> <p>The last requirement is newly introduced in DCASE 2024 Task2.</p> <p> </p> <p><strong>Definition</strong></p> <p>We first define key terms in this task: "machine type," "section," "source domain," "target domain," and "attributes.".</p> <ul> <li>"Machine type" indicates the type of machine, which in the development dataset is one of seven: fan, gearbox, bearing, slide rail, valve, ToyCar, and ToyTrain.</li> <li>A section is defined as a subset of the dataset for calculating performance metrics.</li> <li>The source domain is the domain under which most of the training data and some of the test data were recorded, and the target domain is a different set of domains under which some of the training data and some of the test data were recorded. There are differences between the source and target domains in terms of operating speed, machine load, viscosity, heating temperature, type of environmental noise, signal-to-noise ratio, etc.</li> <li>Attributes are parameters that define states of machines or types of noise. For several machine types, the attributes are hidden.</li> </ul> <p> </p> <p><strong>Dataset</strong></p> <p>This dataset consists of seven machine types. For each machine type, one section is provided, and the section is a complete set of training and test data. For each section, this dataset provides (i) 990 clips of normal sounds in the source domain for training, (ii) ten clips of normal sounds in the target domain for training, and (iii) 100 clips each of normal and anomalous sounds for the test. The source/target domain of each sample is provided. Additionally, the attributes of each sample in the training and test data are provided in the file names and attribute csv files.</p> <p> </p> <p><strong>File names and attribute csv files</strong></p> <p>File names and attribute csv files provide reference labels for each clip. The given reference labels for each training/test clip include machine type, section index, normal/anomaly information, and attributes regarding the condition other than normal/anomaly. The machine type is given by the directory name. The section index is given by their respective file names. For the datasets other than the evaluation dataset, the normal/anomaly information and the attributes are given by their respective file names. Note that for machine types that has its attribute information hidden, the attribute information in each file names are only labeled as "noAttributes". Attribute csv files are for easy access to attributes that cause domain shifts. In these files, the file names, name of parameters that cause domain shifts (domain shift parameter, dp), and the value or type of these parameters (domain shift value, dv) are listed. Each row takes the following format:</p> <p> [filename (string)], [d1p (string)], [d1v (int | float | string)], [d2p], [d2v]...</p> <p>For machine types that have their attribute information hidden, all columns except the filename column are left blank for each row.</p> <p><strong>Recording procedure</strong></p> <p>Normal/anomalous operating sounds of machines and its related equipment are recorded. Anomalous sounds were collected by deliberately damaging target machines. For simplifying the task, we use only the first channel of multi-channel recordings; all recordings are regarded as single-channel recordings of a fixed microphone. We mixed a target machine sound with environmental noise, and only noisy recordings are provided as training/test data. The environmental noise samples were recorded in several real factory environments. We will publish papers on the dataset to explain the details of the recording procedure by the submission deadline.</p> <p> </p> <p><strong>Directory structure</strong></p> <p>- /dev_data </p> <p> - /raw<br> - /fan<br> - /train (only normal clips) <br> - /section_00_source_train_normal_0001_<attribute>.wav <br> - ... <br> - /section_00_source_train_normal_0990_<attribute>.wav <br> - /section_00_target_train_normal_0001_<attribute>.wav <br> - ... <br> - /section_00_target_train_normal_0010_<attribute>.wav <br> - /test <br> - /section_00_source_test_normal_0001_<attribute>.wav <br> - ... <br> - /section_00_source_test_normal_0050_<attribute>.wav <br> - /section_00_source_test_anomaly_0001_<attribute>.wav <br> - ... <br> - /section_00_source_test_anomaly_0050_<attribute>.wav <br> - /section_00_target_test_normal_0001_<attribute>.wav<br> - ... <br> - /section_00_target_test_normal_0050_<attribute>.wav <br> - /section_00_target_test_anomaly_0001_<attribute>.wav <br> - ... <br> - /section_00_target_test_anomaly_0050_<attribute>.wav <br> - attributes_00.csv (attribute csv for section 00)<br> - /gearbox (The other machine types have the same directory structure as fan.) <br> - /bearing<br> - /slider (`slider` means "slide rail")<br> - /ToyCar <br> - /ToyTrain <br> - /valve </p> <p> </p> <p><strong>Baseline system</strong></p> <p>The baseline system is available on the Github repository <<a href="https://github.com/nttcslab/dcase2023_task2_baseline_ae">https://github.com/nttcslab/dcase2023_task2_baseline_ae</a>>.The baseline systems provide a simple entry-level approach that gives a reasonable performance in the dataset of Task 2. They are good starting points, especially for entry-level researchers who want to get familiar with the anomalous-sound-detection task.</p> <p> </p> <p><strong>Condition of use</strong></p> <p>This dataset was created jointly by <strong>Hitachi, Ltd. </strong>and <strong>NTT Corporation</strong> and is available under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license.</p> <p> </p> <p><strong>Citation</strong></p> <p><TBD></p> <p> </p> <p><strong>Contact</strong></p> <p>If there is any problem, please contact us:</p> <ul> <li>Tomoya Nishida, <a href="mailto:kota.dohi.gr@hitachi.com">tomoya.nishida.ax@hitachi.com</a></li> <li>Keisuke Imoto, <a href="mailto:keisuke.imoto@ieee.org">keisuke.imoto@ieee.org</a></li> <li>Noboru Harada, <a href="mailto:noboru@ieee.org">noboru@ieee.org</a></li> <li>Daisuke Niizumi, <a href="mailto:daisuke.niizumi.dt@hco.ntt.co.jp">daisuke.niizumi.dt@hco.ntt.co.jp</a></li> <li>Yohei Kawaguchi, <a href="mailto:yohei.kawaguchi.xk@hitachi.com">yohei.kawaguchi.xk@hitachi.com</a></li> </ul>
Submissions_dcase_2021_task4
<p>Researcher(s)</p> <p>Turpault, Nicola; Salamon, Justin; Wisdom, Scott; Erdogan, Hakan; Hershey, John; Seetharaman, Prem; Ellis, Daniel P. W; Cornell, Samuele; Fonseca, Eduardo. </p> <p>Predictions and technical reports of the systems submitted to DCASE task 4 2021.</p> <p>Mapping files and ground-truth related to the different versions of the synthetic evaluation datasets used in [1] are also available in the ICASSP_folder.zip file.</p> <p><br> [1] Ronchini F., Serizel R., “A benchmark of state-of-the-art sound event detection systems<br> evaluated on synthetic soundscapes”, in ICASSP 2022 IEEE International Conference on<br> Acoustics, Speech and Signal Processing (ICASSP)</p> <p> </p>
Test_S: a synthesised dataset for evaluating few-shot Bioacoustic events detection system (DCASE 2021 Task5)
<ul> <li>The database consists of : <ul> <li>DC class and MEL class </li> <li>dense bioacoustic event like official Evaluation Set in each audio</li> </ul> </li> </ul>
DCASE 2022 Challenge Task 2 Evaluation Dataset
<p><strong>Description</strong></p> <p>This dataset is the "evaluation dataset" for the <a href="https://dcase.community/challenge2022/task-unsupervised-anomalous-sound-detection-for-machine-condition-monitoring"><strong>DCASE 2022 Challenge Task 2 "Unsupervised Anomalous Sound Detection for Machine Condition Monitoring Applying Domain Generalization Techniques</strong>"</a>.</p> <p> </p> <p><strong>Condition of use</strong></p> <p>This dataset was created jointly by <strong>Hitachi, Ltd. </strong>and <strong>NTT Corporation</strong> and is available under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license.</p> <p> </p> <p><strong>Citation</strong></p> <p>If you use this dataset, please cite all the following three papers. </p> <ul> <li>Kota Dohi, Keisuke Imoto, Noboru Harada, Daisuke Niizumi, Yuma Koizumi, Tomoya Nishida, Harsh Purohit, Takashi Endo, Masaaki Yamamoto, Yohei Kawaguchi, <em>Description and Discussion on DCASE 2022 Challenge Task 2: Unsupervised Anomalous Sound Detection for Machine Condition Monitoring Applying Domain Generalization Techniques. In arXiv e-prints: 2206.05876, </em>2022. [<a href="https://arxiv.org/abs/2206.05876">URL</a>]</li> <li>Kota Dohi, Tomoya Nishida, Harsh Purohit, Ryo Tanabe, Takashi Endo, Masaaki Yamamoto, Yuki Nikaido, and Yohei Kawaguchi. <em>MIMII DG: sound dataset for malfunctioning industrial machine investigation and inspection for domain generalization task.</em> <em>In arXiv e-prints: 2205.13879</em>, 2022. [<a href="https://arxiv.org/pdf/2205.13879.pdf">URL</a>]</li> <li>Noboru Harada, Daisuke Niizumi, Daiki Takeuchi, Yasunori Ohishi, Masahiro Yasuda, and Shoichiro Saito. <em>ToyADMOS2: another dataset of miniature-machine operating sounds for anomalous sound detection under domain shift conditions.</em> In Proceedings of the 6th Detection and Classification of Acoustic Scenes and Events 2021 Workshop (DCASE2021), 1–5. Barcelona, Spain, November 2021. [<a href="https://dcase.community/documents/workshop2021/proceedings/DCASE2021Workshop_Harada_6.pdf">URL</a>]</li> </ul> <p><strong>Contact</strong></p> <p>If there is any problem, please contact us:</p> <ul> <li>Kota Dohi, <a href="mailto:kota.dohi.gr@hitachi.com">kota.dohi.gr@hitachi.com</a></li> <li>Daisuke Niizumi, <a href="mailto:daisuke.niizumi.dt@hco.ntt.co.jp">daisuke.niizumi.dt@hco.ntt.co.jp</a></li> <li>Yohei Kawaguchi, <a href="mailto:yohei.kawaguchi.xk@hitachi.com">yohei.kawaguchi.xk@hitachi.com</a></li> <li>Keisuke Imoto, <a href="mailto:keisuke.imoto@ieee.org">keisuke.imoto@ieee.org</a></li> </ul>
DCASE 2024 Challenge Task 2 Additional Training Dataset
<p><strong><Important ! Notes on releasing Version v2 (23 May, 2024)></strong><br><strong>Due to some data issues, data files for 3DPrinter and RoboticArm have been updated. The new versions of these files are renamed as "eval_data_3DPrinter_train_r2.zip" and "eval_data_RoboticArm_train_r2.zip". Please use these files for the DCASE 2024 Challenge Task 2. (Other files have not been changed from Version v1) We apologize for your inconvenience.</strong></p> <p><strong>Description</strong></p> <p>This dataset is the "additional training dataset" for the <a href="https://dcase.community/challenge2024/task-first-shot-unsupervised-anomalous-sound-detection-for-machine-condition-monitoring">DCASE 2024 Challenge Task 2 "First-Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring"</a>.</p> <p>The data consists of the normal/anomalous operating sounds of nine types of real/toy machines. Each recording is a single-channel audio that includes both a machine's operating sound and environmental noise. The duration of recordings varies from 6 to 10 seconds. The following nine types of real/toy machines are used in this task:</p> <ul> <li>3DPrinter</li> <li>AirCompressor</li> <li>BrushlessMotor</li> <li>HairDryer</li> <li>HoveringDrone</li> <li>RoboticArm</li> <li>Scanner</li> <li>ToothBrush</li> <li>ToyCircuit</li> </ul> <p><strong>Overview of the task</strong></p> <p><strong>Anomalous sound detection (ASD) is the task of identifying whether the sound emitted from a target machine is normal or anomalous. </strong>Automatic detection of mechanical failure is an essential technology in the fourth industrial revolution, which involves artificial-intelligence-based factory automation. Prompt detection of machine anomalies by observing sounds is useful for monitoring the condition of machines. </p> <p>This task is the follow-up from DCASE 2020 Task 2 to DCASE 2023 Task 2. The task this year is to develop an ASD system that meets the following five requirements.</p> <p><strong>1. Train a model using only normal sound (unsupervised learning scenario)</strong> <br>Because anomalies rarely occur and are highly diverse in real-world factories, it can be difficult to collect exhaustive patterns of anomalous sounds. Therefore, the system must detect unknown types of anomalous sounds that are not provided in the training data. This is the same requirement as in the previous tasks.</p> <p><strong>2. Detect anomalies regardless of domain shifts (domain generalization task) </strong><br>In real-world cases, the operational states of a machine or the environmental noise can change to cause domain shifts. Domain-generalization techniques can be useful for handling domain shifts that occur frequently or are hard-to-notice. In this task, the system is required to use domain-generalization techniques for handling these domain shifts. This requirement is the same as in DCASE 2022 Task 2 and DCASE 2023 Task 2.</p> <p><strong>3. Train a model for a completely new machine type</strong><br>For a completely new machine type, hyperparameters of the trained model cannot be tuned. Therefore, the system should have the ability to train models without additional hyperparameter tuning. This requirement is the same as in DCASE 2023 Task 2.</p> <p><strong>4. Train a model using a limited number of machines from its machine type</strong><br>While sounds from multiple machines of the same machine type can be used to enhance the detection performance, it is often the case that only a limited number of machines are available for a machine type. In such a case, the system should be able to train models using a few machines from a machine type. This requirement is the same as in DCASE 2023 Task 2.</p> <p><strong>5 . Train a model both with or without attribute information</strong><br>While additional attribute information can help enhance the detection performance, we cannot always obtain such information. Therefore, the system must work well both when attribute information is available and when it is not.</p> <p>The last requirement is newly introduced in DCASE 2024 Task2.</p> <p> </p> <p><strong>Definition</strong></p> <p>We first define key terms in this task: "machine type," "section," "source domain," "target domain," and "attributes.".</p> <ul> <li>"Machine type" indicates the type of machine, which in the additional training dataset is one of nine: 3D-printer, air compressor, brushless motor, hair dryer, hovering drone, robotic arm, document scanner (scanner), toothbrush, and Toy circuit.</li> <li>A section is defined as a subset of the dataset for calculating performance metrics.</li> <li>The source domain is the domain under which most of the training data and some of the test data were recorded, and the target domain is a different set of domains under which some of the training data and some of the test data were recorded. There are differences between the source and target domains in terms of operating speed, machine load, viscosity, heating temperature, type of environmental noise, signal-to-noise ratio, etc.</li> <li>Attributes are parameters that define states of machines or types of noise. For several machine types, the attributes are hidden.</li> </ul> <p> </p> <p><strong>Dataset</strong></p> <p>This dataset consists of nine machine types. For each machine type, one section is provided, and the section is a complete set of training data. A set of test data corresponding to this training data will be provided in another seperate zenodo page as an "evaluation dataset" for the DCASE 2024 Challenge task 2. For each section, this dataset provides (i) 990 clips of normal sounds in the source domain for training and (ii) ten clips of normal sounds in the target domain for training. The source/target domain of each sample is provided. Additionally, the attributes of each sample in the training and test data are provided in the file names and attribute csv files.</p> <p> </p> <p><strong>File names and attribute csv files</strong></p> <p>File names and attribute csv files provide reference labels for each clip. The given reference labels for each training clip include machine type, section index, normal/anomaly information, and attributes regarding the condition other than normal/anomaly. The machine type is given by the directory name. The section index is given by their respective file names. For the datasets other than the evaluation dataset, the normal/anomaly information and the attributes are given by their respective file names. Note that for machine types that has its attribute information hidden, the attribute information in each file names are only labeled as "noAttributes". Attribute csv files are for easy access to attributes that cause domain shifts. In these files, the file names, name of parameters that cause domain shifts (domain shift parameter, dp), and the value or type of these parameters (domain shift value, dv) are listed. Each row takes the following format:</p> <p> [filename (string)], [d1p (string)], [d1v (int | float | string)], [d2p], [d2v]...</p> <p>For machine types that have their attribute information hidden, all columns except the filename column are left blank for each row.</p> <p><strong>Recording procedure</strong></p> <p>Normal/anomalous operating sounds of machines and its related equipment are recorded. Anomalous sounds were collected by deliberately damaging target machines. For simplifying the task, we use only the first channel of multi-channel recordings; all recordings are regarded as single-channel recordings of a fixed microphone. We mixed a target machine sound with environmental noise, and only noisy recordings are provided as training/test data. The environmental noise samples were recorded in several real factory environments. We will publish papers on the dataset to explain the details of the recording procedure by the submission deadline.</p> <p> </p> <p><strong>Directory structure</strong></p> <p>- /eval_data </p> <p> - /raw<br> - /3DPrinter<br> - /train (only normal clips) <br> - /section_00_source_train_normal_0001_<attribute>.wav <br> - ... <br> - /section_00_source_train_normal_0990_<attribute>.wav <br> - /section_00_target_train_normal_0001_<attribute>.wav <br> - ... <br> - /section_00_target_train_normal_0010_<attribute>.wav <br> - attributes_00.csv (attribute csv for section 00)<br> - /AirCompressor (The other machine types have the same directory structure as 3DPrinter.) <br> - /BrushlessMotor<br> - /HairDryer<br> - /HoveringDrone<br> - /RoboticArm<br> - /Scanner<br> - /ToothBrush<br> - /ToyCircuit</p> <p> </p> <p> </p> <p><strong>Baseline system</strong></p> <p>The baseline system is available on the Github repository <<a href="https://github.com/nttcslab/dcase2023_task2_baseline_ae">https://github.com/nttcslab/dcase2023_task2_baseline_ae</a>>. The baseline systems provide a simple entry-level approach that gives a reasonable performance in the dataset of Task 2. They are good starting points, especially for entry-level researchers who want to get familiar with the anomalous-sound-detection task.</p> <p> </p> <p><strong>Condition of use</strong></p> <p>This dataset was created jointly by <strong>Hitachi, Ltd.,</strong> <strong>NTT Corporation and STMicroelectronics</strong> and is available under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license.</p> <p> </p> <p><strong>Citation</strong></p> <p><TBD></p> <p> </p> <p><strong>Contact</strong></p> <p>If there is any problem, please contact us:</p> <ul> <li>Tomoya Nishida, <a href="mailto:kota.dohi.gr@hitachi.com">tomoya.nishida.ax@hitachi.com</a></li> <li>Keisuke Imoto, <a href="mailto:keisuke.imoto@ieee.org">keisuke.imoto@ieee.org</a></li> <li>Noboru Harada, <a href="mailto:noboru@ieee.org">noboru@ieee.org</a></li> <li>Daisuke Niizumi, <a href="mailto:daisuke.niizumi.dt@hco.ntt.co.jp">daisuke.niizumi.dt@hco.ntt.co.jp</a></li> <li>Yohei Kawaguchi, <a href="mailto:yohei.kawaguchi.xk@hitachi.com">yohei.kawaguchi.xk@hitachi.com</a></li> </ul>
Evaluation set DCASE 2024 task 4 (for submissions)
<p>This repo contains the evaluation dataset to download and submit system outputs for when participating in task 4 of the DCASE 2024 Challenge.</p>
Ground Truth for DCASE 2021 Challenge Task 2 Evaluation Dataset
<p><strong>Description</strong></p> <p>This data is the ground truth for the "<a href="https://zenodo.org/record/4884786">evaluation dataset</a>" for the <a href="http://dcase.community/challenge2021/task-unsupervised-detection-of-anomalous-sounds"><strong>DCASE 2021 Challenge Task 2 "Unsupervised Anomalous Sound Detection for Machine Condition Monitoring under Domain Shifted Conditions"</strong></a>. </p> <p>In the task, three datasets have been released: "<a href="http://zenodo.org/record/4562016">development dataset</a>", "<a href="https://zenodo.org/record/4660992">additional training dataset</a>", and "<a href="https://zenodo.org/record/4884786">evaluation dataset</a>". The evaluation dataset was the last of the three released and includes around 200 samples for each machine type, section index, and domain, none of which have a condition label (i.e., normal or anomaly). This ground truth dataset contains the condition labels.</p> <p> </p> <p><strong>Data format</strong></p> <p>The CSV file for each machine type, section index, and domain includes the ground truth data like the following:</p> <p>---------------------------------</p> <p>section_03_source_test_0000.wav,1<br> section_03_source_test_0001.wav,1</p> <p>...</p> <p>section_03_source_test_0198.wav,0<br> section_03_source_test_0199.wav,1</p> <p>---------------------------------</p> <p>The first column shows the name of a wave file. The second column shows the condition label (i.e., 0: normal or 1: anomaly).</p> <p> </p> <p><strong>How to use</strong></p> <p>A script for calculating the AUC, pAUC, precision, recall, and F1 scores for the "evaluation dataset" is available on the Github repository <a href="https://github.com/y-kawagu/dcase2021_task2_evaluator">[URL]</a>. The ground truth data are used by this system. For more information, please see the Github repository.</p> <p> </p> <p><strong>Conditions of use</strong></p> <p>This dataset was created jointly by <strong>Hitachi, Ltd.</strong> and <strong>NTT Corporation</strong> and is available under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license.</p> <p> </p> <p><strong>Publication</strong></p> <p>If you use this dataset, please cite <strong>all the following three papers</strong>:</p> <ul> <li>Yohei Kawaguchi, Keisuke Imoto, Yuma Koizumi, Noboru Harada, Daisuke Niizumi, Kota Dohi, Ryo Tanabe, Harsh Purohit, and Takashi Endo, "Description and Discussion on DCASE 2021 Challenge Task 2: Unsupervised Anomalous Sound Detection for Machine Condition Monitoring under Domain Shifted Conditions," in arXiv e-prints: 2106.04492, 2021. [<a href="https://arxiv.org/abs/2106.04492">URL</a>]</li> <li>Noboru Harada, Daisuke Niizumi, Daiki Takeuchi, Yasunori Ohishi, Masahiro Yasuda, Shoichiro Saito, "ToyADMOS2: Another Dataset of Miniature-Machine Operating Sounds for Anomalous Sound Detection under Domain Shift Conditions," in arXiv e-prints: 2106.02369, 2021. [<a href="https://arxiv.org/abs/2106.02369">URL</a>]</li> <li>Ryo Tanabe, Harsh Purohit, Kota Dohi, Takashi Endo, Yuki Nikaido, Toshiki Nakamura, and Yohei Kawaguchi, "MIMII DUE: Sound Dataset for Malfunctioning Industrial Machine Investigation and Inspection with Domain Shifts due to Changes in Operational and Environmental Conditions," in arXiv e-prints: 2105.02702, 2021. [<a href="https://arxiv.org/abs/2105.02702">URL</a>]</li> </ul> <p><br> <strong>Feedback</strong></p> <p>If there is any problem, please contact us:</p> <ul> <li>Yohei Kawaguchi, <a href="mailto:yohei.kawaguchi.xk@hitachi.com">yohei.kawaguchi.xk@hitachi.com</a></li> <li>Daisuke Niizumi, <a href="mailto:daisuke.niizumi.dt@hco.ntt.co.jp">daisuke.niizumi.dt@hco.ntt.co.jp</a></li> <li>Keisuke Imoto, <a href="mailto:keisuke.imoto@ieee.org">keisuke.imoto@ieee.org</a></li> </ul> <p> </p>
DCASE 2023 Challenge Task 2 Evaluation Dataset
<p><strong>Description</strong></p> <p>This dataset is the "evaluation dataset" for the <a href="https://dcase.community/challenge2023/task-first-shot-unsupervised-anomalous-sound-detection-for-machine-condition-monitoring">DCASE 2023 Challenge Task 2 "First-Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring"</a>.</p> <p>The data consists of the normal/anomalous operating sounds of seven types of real/toy machines. Each recording is a single-channel audio that includes both a machine's operating sound and environmental noise. The duration of recordings varies from 6 to 18 sec, depending on the machine type. The following seven types of real/toy machines are used:</p> <ul> <li>Vacuum</li> <li>ToyTank</li> <li>ToyNscale</li> <li>ToyDrone</li> <li>bandsaw</li> <li>grinder</li> <li>shaker</li> </ul> <p> </p> <p><strong>Definition</strong></p> <p>We first define key terms in this task: "machine type," "section," "source domain," "target domain," and "attributes.".</p> <ul> <li>"Machine type" indicates the type of machine, which in the development dataset is one of seven: fan, gearbox, bearing, slide rail, valve, ToyCar, and ToyTrain.</li> <li>A section is defined as a subset of the dataset for calculating performance metrics.</li> <li>The source domain is the domain under which most of the training data and some of the test data were recorded, and the target domain is a different set of domains under which some of the training data and some of the test data were recorded. There are differences between the source and target domains in terms of operating speed, machine load, viscosity, heating temperature, type of environmental noise, signal-to-noise ratio, etc.</li> <li>Attributes are parameters that define states of machines or types of noise.</li> </ul> <p> </p> <p><strong>Dataset</strong></p> <p>This dataset consists of seven machine types. For each machine type, one section is provided, and the section is a complete set of training and test data. For each section, this dataset provides (i) 990 clips of normal sounds in the source domain for training, (ii) ten clips of normal sounds in the target domain for training. The source/target domain of each sample is provided. Additionally, the attributes of each sample in the training and test data are provided in the file names and attribute csv files.</p> <p> </p> <p><strong>Recording procedure</strong></p> <p>Normal/anomalous operating sounds of machines and its related equipment are recorded. Anomalous sounds were collected by deliberately damaging target machines. For simplifying the task, we use only the first channel of multi-channel recordings; all recordings are regarded as single-channel recordings of a fixed microphone. We mixed a target machine sound with environmental noise, and only noisy recordings are provided as training/test data. The environmental noise samples were recorded in several real factory environments. We will publish papers on the dataset to explain the details of the recording procedure by the submission deadline.</p> <p> </p> <p><strong>Directory structure</strong></p> <p>- /dev_data </p> <p> - /raw<br> - /Vacuum<br> - /test <br> - /section_00_0001.wav <br> - ... <br> - /section_00_0200.wav <br> - /ToyTank (The other machine types have the same directory structure as Vacuum.) <br> - /ToyNscale<br> - /ToyDrone<br> - /bandsaw<br> - /grinder<br> - /shaker</p> <p><strong>Condition of use</strong></p> <p>This dataset was created jointly by <strong>Hitachi, Ltd. </strong>and <strong>NTT Corporation</strong> and is available under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license.</p> <p> </p> <p><strong>Citation</strong></p> <p>If you use this dataset, please cite all the following papers. We will publish a paper on the description of the DCASE 2023 Task 2, so pleasure make sure to cite the paper, too.</p> <ul> <li>Noboru Harada, Daisuke Niizumi, Yasunori Ohishi, Daiki Takeuchi, and Masahiro Yasuda. <em>First-shot anomaly detection for machine condition monitoring: A domain generalization baseline. In arXiv e-prints: 2303.00455</em>, 2023. [<a href="https://arxiv.org/pdf/2303.00455.pdf">URL</a>]</li> <li>Kota Dohi, Tomoya Nishida, Harsh Purohit, Ryo Tanabe, Takashi Endo, Masaaki Yamamoto, Yuki Nikaido, and Yohei Kawaguchi. <em>MIMII DG: sound dataset for malfunctioning industrial machine investigation and inspection for domain generalization task.</em> In Proceedings of the 7th Detection and Classification of Acoustic Scenes and Events 2022 Workshop (DCASE2022), 31-35. Nancy, France, November 2022, . [<a href="https://arxiv.org/pdf/2205.13879.pdf">URL</a>]</li> <li>Noboru Harada, Daisuke Niizumi, Daiki Takeuchi, Yasunori Ohishi, Masahiro Yasuda, and Shoichiro Saito. <em>ToyADMOS2: another dataset of miniature-machine operating sounds for anomalous sound detection under domain shift conditions.</em> In Proceedings of the 6th Detection and Classification of Acoustic Scenes and Events 2021 Workshop (DCASE2021), 1–5. Barcelona, Spain, November 2021. [<a href="https://dcase.community/documents/workshop2021/proceedings/DCASE2021Workshop_Harada_6.pdf">URL</a>]</li> </ul> <p> </p> <p><strong>Contact</strong></p> <p>If there is any problem, please contact us:</p> <ul> <li>Kota Dohi, <a href="mailto:kota.dohi.gr@hitachi.com">kota.dohi.gr@hitachi.com</a></li> <li>Keisuke Imoto, <a href="mailto:keisuke.imoto@ieee.org">keisuke.imoto@ieee.org</a></li> <li>Noboru Harada, <a href="mailto:noboru@ieee.org">noboru@ieee.org</a></li> <li>Daisuke Niizumi, <a href="mailto:daisuke.niizumi.dt@hco.ntt.co.jp">daisuke.niizumi.dt@hco.ntt.co.jp</a></li> <li>Yohei Kawaguchi, <a href="mailto:yohei.kawaguchi.xk@hitachi.com">yohei.kawaguchi.xk@hitachi.com</a></li> </ul>
DCASE 2024 Challenge Task 2 Evaluation Dataset
<p><strong><Important ! Version v2 of "DCASE 2024 Challenge Task 2 Additional Training dataset" has been published on 23 May, 2024></strong><br><strong>Version v2 of the "DCASE 2024 Challenge Task 2 Additional Training dataset", which corresponds to this "DCASE 2024 Challenge Task 2 Evaluation Dataset (Version v1)", has been published on <a href="../records/11259435">https://zenodo.org/records/11259435</a> (23 May, 2024). In the "DCASE 2024 Challenge Task 2 Additional Training dataset", data files for 3DPrinter and RoboticArm are updated from Version v1 and are renamed as "eval_data_3DPrinter_train_r2.zip" and "eval_data_RoboticArm_train_r2.zip". Please use those files for the "Additional Training dataset" in DCASE 2024 Challenge Task 2. (Other files have not been changed from Version v1.) We apologize for your inconvenience.</strong></p> <p> </p> <p><strong><The data files will be made publicly accessible on June 1st, 2024></strong></p> <p><strong>Description</strong></p> <p>This dataset is the "evaluation dataset" for the <a href="https://dcase.community/challenge2024/task-first-shot-unsupervised-anomalous-sound-detection-for-machine-condition-monitoring">DCASE 2024 Challenge Task 2 "First-Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring"</a>.</p> <p>The data consists of the normal/anomalous operating sounds of nine types of real/toy machines. Each recording is a single-channel audio that includes both a machine's operating sound and environmental noise. The duration of recordings varies from 6 to 10 seconds. The following nine types of real/toy machines are used in this task:</p> <ul> <li>3DPrinter</li> <li>AirCompressor</li> <li>BrushlessMotor</li> <li>HairDryer</li> <li>HoveringDrone</li> <li>RoboticArm</li> <li>Scanner</li> <li>ToothBrush</li> <li>ToyCircuit</li> </ul> <p> </p> <p><strong>Overview of the task</strong></p> <p><strong>Anomalous sound detection (ASD) is the task of identifying whether the sound emitted from a target machine is normal or anomalous. </strong>Automatic detection of mechanical failure is an essential technology in the fourth industrial revolution, which involves artificial-intelligence-based factory automation. Prompt detection of machine anomalies by observing sounds is useful for monitoring the condition of machines. </p> <p>This task is the follow-up from DCASE 2020 Task 2 to DCASE 2023 Task 2. The task this year is to develop an ASD system that meets the following five requirements.</p> <p><strong>1. Train a model using only normal sound (unsupervised learning scenario)</strong> <br>Because anomalies rarely occur and are highly diverse in real-world factories, it can be difficult to collect exhaustive patterns of anomalous sounds. Therefore, the system must detect unknown types of anomalous sounds that are not provided in the training data. This is the same requirement as in the previous tasks.</p> <p><strong>2. Detect anomalies regardless of domain shifts (domain generalization task) </strong><br>In real-world cases, the operational states of a machine or the environmental noise can change to cause domain shifts. Domain-generalization techniques can be useful for handling domain shifts that occur frequently or are hard-to-notice. In this task, the system is required to use domain-generalization techniques for handling these domain shifts. This requirement is the same as in DCASE 2022 Task 2 and DCASE 2023 Task 2.</p> <p><strong>3. Train a model for a completely new machine type</strong><br>For a completely new machine type, hyperparameters of the trained model cannot be tuned. Therefore, the system should have the ability to train models without additional hyperparameter tuning. This requirement is the same as in DCASE 2023 Task 2.</p> <p><strong>4. Train a model using a limited number of machines from its machine type</strong><br>While sounds from multiple machines of the same machine type can be used to enhance the detection performance, it is often the case that only a limited number of machines are available for a machine type. In such a case, the system should be able to train models using a few machines from a machine type. This requirement is the same as in DCASE 2023 Task 2.</p> <p><strong>5 . Train a model both with or without attribute information</strong><br>While additional attribute information can help enhance the detection performance, we cannot always obtain such information. Therefore, the system must work well both when attribute information is available and when it is not.</p> <p>The last requirement is newly introduced in DCASE 2024 Task2.</p> <p> </p> <p><strong>Definition</strong></p> <p>We first define key terms in this task: "machine type," "section," "source domain," "target domain," and "attributes.".</p> <ul> <li>"Machine type" indicates the type of machine, which in the additional training dataset is one of nine: 3D-printer, air compressor, brushless motor, hair dryer, hovering drone, robotic arm, document scanner (scanner), toothbrush, and Toy circuit.</li> <li>A section is defined as a subset of the dataset for calculating performance metrics.</li> <li>The source domain is the domain under which most of the training data and some of the test data were recorded, and the target domain is a different set of domains under which some of the training data and some of the test data were recorded. There are differences between the source and target domains in terms of operating speed, machine load, viscosity, heating temperature, type of environmental noise, signal-to-noise ratio, etc.</li> <li>Attributes are parameters that define states of machines or types of noise. For several machine types, the attributes are hidden.</li> </ul> <p> </p> <p><strong>Dataset</strong></p> <p>This dataset consists of nine machine types. For each machine type, one section is provided, and the section is a complete set of test data. A set of training data corresponding to this test data is provided in another seperate zenodo page as an "additional training dataset" for the DCASE 2024 Challenge task 2 (<a href="../records/11259435">https://zenodo.org/records/11259435</a>). For each section, this dataset provides 200 clips of test data.</p> <p> </p> <p><strong>Recording procedure</strong></p> <p>Normal/anomalous operating sounds of machines and its related equipment are recorded. Anomalous sounds were collected by deliberately damaging target machines. For simplifying the task, we use only the first channel of multi-channel recordings; all recordings are regarded as single-channel recordings of a fixed microphone. We mixed a target machine sound with environmental noise, and only noisy recordings are provided as training/test data. The environmental noise samples were recorded in several real factory environments. We will publish papers on the dataset to explain the details of the recording procedure by the submission deadline.</p> <p> </p> <p><strong>Directory structure</strong></p> <p>- /eval_data </p> <p> - /raw<br> - /3DPrinter<br> - /test <br> - /section_00_0001.wav <br> - ... <br> - /section_00_0200.wav <br> - /AirCompressor (The other machine types have the same directory structure as 3DPrinter.) <br> - /BrushlessMotor<br> - /HairDryer<br> - /HoveringDrone<br> - /RoboticArm<br> - /Scanner<br> - /ToothBrush<br> - /ToyCircuit</p> <p> </p> <p><strong>Baseline system</strong></p> <p>The baseline system is available on the Github repository <<a href="https://github.com/nttcslab/dcase2023_task2_baseline_ae">https://github.com/nttcslab/dcase2023_task2_baseline_ae</a>>. The baseline systems provide a simple entry-level approach that gives a reasonable performance in the dataset of Task 2. They are good starting points, especially for entry-level researchers who want to get familiar with the anomalous-sound-detection task.</p> <p> </p> <p><strong>Condition of use</strong></p> <p>This dataset was created jointly by <strong>Hitachi, Ltd.,</strong> <strong>NTT Corporation and STMicroelectronics</strong> and is available under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license.</p> <p> </p> <p><strong>Citation</strong></p> <p><TBD></p> <p> </p> <p><strong>Contact</strong></p> <p>If there is any problem, please contact us:</p> <ul> <li>Tomoya Nishida, <a href="mailto:kota.dohi.gr@hitachi.com">tomoya.nishida.ax@hitachi.com</a></li> <li>Keisuke Imoto, <a href="mailto:keisuke.imoto@ieee.org">keisuke.imoto@ieee.org</a></li> <li>Noboru Harada, <a href="mailto:noboru@ieee.org">noboru@ieee.org</a></li> <li>Daisuke Niizumi, <a href="mailto:daisuke.niizumi.dt@hco.ntt.co.jp">daisuke.niizumi.dt@hco.ntt.co.jp</a></li> <li>Yohei Kawaguchi, <a href="mailto:yohei.kawaguchi.xk@hitachi.com">yohei.kawaguchi.xk@hitachi.com</a></li> </ul>
Task_4_DCASE_2017_training_set
<p>The dataset from Google Drive</p>
DCASE 2024 Challenge Task 10 Development Dataset: Acoustic-based Traffic Monitoring
<h3><strong>Directory structure:</strong></h3> <p><strong>engine-sounds.zip</strong><br>|----- car [<em>car engine sounds</em>]<br>|----- cv [<em>commercial vehicles engine sounds</em>]</p> <p><strong>locX.zip</strong><br>|----- meta.json [<em>contains meta information of traffic condition and sensor setup corresponding to the location</em>]<br>|----- train [<em>train flac files inside</em>]<br>|----- train.csv<br>|----- val [<em>val flac files inside</em>]<br>|----- val.csv</p> <p><strong>simulation.zip</strong><br>|----- locX<br> |----- car<br> | |_____ left [<em>flac files and label csv inside</em>]<br> | |_____ right [<em>flac files and label csv inside</em>]<br> |----- cv<br> |_____ left [<em>flac files and label csv inside</em>]<br> |_____ right [<em>flac files and label csv inside</em>]</p> <p><strong>NOTE:</strong> We split zip files for large size folders (i.e., loc1, loc3, loc6). Please make sure to download <em><strong>all splits</strong></em> before unzipping the data! For example, run </p> <p><code>zip -s 0 loc1.zip --out unsplit-loc1.zip</code></p> <p><code>unzip unsplit-loc1.zip</code></p> <p>in your terminal once you have downloaded <code>loc1.z01</code>, <code>loc1.z02</code>, <code>loc1.z03</code> and <code>loc1.zip</code>.</p> <p> </p> <h3><strong>Contact:</strong></h3> <p>If there is any question, please contact:</p> <ul> <li>Luca Bondi: <a href="mailto:luca.bondi@us.bosch.com">luca </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:luca.bondi@us.bosch.com"> bondi@us (dot) bosch (dot) com</a></li> <li>Shabnam Ghaffarzadegan: <a href="mailto:shabnam.ghaffarzadegan@us.bosch.com">shabnam </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:shabnam.ghaffarzadegan@us.bosch.com"> ghaffarzadegan@us </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:shabnam.ghaffarzadegan@us.bosch.com"> bosch </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:shabnam.ghaffarzadegan@us.bosch.com"> com</a></li> <li>Wei-Cheng (Winston) Lin: <a href="mailto:winston.lin@us.bosch.com">wei-cheng </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:winston.lin@us.bosch.com"> lin@us </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:winston.lin@us.bosch.com"> bosch </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:winston.lin@us.bosch.com"> com</a></li> </ul> <p> </p> <h3><strong>Acknowledgement:</strong></h3> <p><em>This work has partially received funding from the European Union's Horizon 2020 research and innovation programme under the Marie </em><em>Skłodowska</em><em>-Curie grant agreement No. 956962 and from the European Research Council under the European Union's Horizon 2020 research and innovation program / ERC Consolidator Grant: SONORA (no. 773268). This work reflects only the authors' views and the Union is not liable for any use that may be made of the contained information.</em></p> <p> </p>
DCASE 2024 Challenge Task 10 Evaluation Dataset: Acoustic-based Traffic Monitoring
<h3><strong>Directory structure:</strong></h3> <p><strong>locX.zip</strong><br>|----- test [<em>test flac files inside</em>]</p> <p> </p> <p><strong>NOTE:</strong> We split zip files for large size folders (i.e., loc1, loc3, loc6). Please make sure to download <em><strong>all splits</strong></em> before unzipping the data! For example, run </p> <p><code>zip -s 0 loc6.zip --out unsplit-loc6.zip</code></p> <p><code>unzip unsplit-loc6.zip</code></p> <p>in your terminal once you have downloaded <code>loc6.z01</code>, <code>loc6.z02</code> and <code>loc6.zip</code>.</p> <p> </p> <h3><strong>Contact:</strong></h3> <p>If there is any question, please contact:</p> <ul> <li>Luca Bondi: <a href="mailto:luca.bondi@us.bosch.com">luca </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:luca.bondi@us.bosch.com"> bondi@us (dot) bosch (dot) com</a></li> <li>Shabnam Ghaffarzadegan: <a href="mailto:shabnam.ghaffarzadegan@us.bosch.com">shabnam </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:shabnam.ghaffarzadegan@us.bosch.com"> ghaffarzadegan@us </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:shabnam.ghaffarzadegan@us.bosch.com"> bosch </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:shabnam.ghaffarzadegan@us.bosch.com"> com</a></li> <li>Wei-Cheng (Winston) Lin: <a href="mailto:winston.lin@us.bosch.com">wei-cheng </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:winston.lin@us.bosch.com"> lin@us </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:winston.lin@us.bosch.com"> bosch </a><a href="mailto:luca.bondi@us.bosch.com">(dot)</a><a href="mailto:winston.lin@us.bosch.com"> com</a></li> </ul> <p> </p> <h3><strong>Acknowledgement:</strong></h3> <p><em>This work has partially received funding from the European Union's Horizon 2020 research and innovation programme under the Marie </em><em>Skłodowska</em><em>-Curie grant agreement No. 956962 and from the European Research Council under the European Union's Horizon 2020 research and innovation program / ERC Consolidator Grant: SONORA (no. 773268). This work reflects only the authors' views and the Union is not liable for any use that may be made of the contained information.</em></p>
ScienceDex guides
Understand access before you commit
These curated guides explain access requirements, typical timelines, costs, and reuse considerations for widely used research datasets.
Allen Brain Atlas
Allen Brain Atlas is an Allen Institute collection of brain map atlases, datasets, APIs, and analysis tools covering mouse, human, and non-human primate brain resources.
Annotated Behaviour and Observability Dataset (ABODe)
ABODe is a University of Edinburgh DataShare dataset for behavior classification in group-housed mice using home-cage video, identities, bounding boxes, ground-plate positions, and annotator labels.
DANDI Archive for NWB datasets
DANDI is a BRAIN Initiative archive for publishing and sharing neurophysiology data, including electrophysiology, optophysiology, and behavioral data packaged as NWB and related standards.
International Brain Laboratory public data
The International Brain Laboratory public data releases expose standardized mouse decision-making experiments, including Neuropixels recordings, widefield calcium imaging, behavior, and session metadata accessed through the ONE API.
OpenNeuro
OpenNeuro is a free, open platform for sharing neuroimaging datasets, with public search, dataset pages, and download paths for web, S3, DataLad, and the OpenNeuro CLI.