19,997 machine learning datasets
19,997 dataset results
Despite its importance for assessing the effectiveness of communicating information visually, fine-grained recallability of information visualisations has not been studied quantitatively so far.
Phrase in Context is a curated benchmark for phrase understanding and semantic search, consisting of three tasks of increasing difficulty: Phrase Similarity (PS), Phrase Retrieval (PR) and Phrase Sense Disambiguation (PSD). The datasets are annotated by 13 linguistic experts on Upwork and verified by two groups: ~1000 AMT crowdworkers and another set of 5 linguistic experts. PiC benchmark is distributed under CC-BY-NC 4.0.
This data is derived from the 7Scenes dataset. It contains graphs used for training PoserNet and for evaluating its performance.
MNIST Multiview Datasets MNIST is a publicly available dataset consisting of 70, 000 images of handwritten digits distributed over ten classes. We generated 2 four-view datasets where each view is a vector of R<sup>14 x 14</sup>:
The HELMET dataset contains 910 videoclips of motorcycle traffic, recorded at 12 observation sites in Myanmar in 2016. Each videoclip has a duration of 10 seconds, recorded with a framerate of 10fps and a resolution of 1920x1080. The dataset contains 10,006 individual motorcycles, surpassing the number of motorcycles available in existing datasets. Each motorcycle in the 91,000 annotated frames of the dataset is annotated with a bounding box, and rider number per motorcycle as well as position specific helmet use data is available.
Manually vAlidated Vq2a Examples fRom Image/Caption datasetS (MAVERICS) is a suite of test-only visual question answering datasets.
Main Dataset city_pollution_data.csv
The Western Mediterranean Wetlands Bird Dataset is a collection of birds' vocalizations of different lengths that primarily consists of 5,795 labelled audio clips derived from 1,098 recordings, totalling 201.6 minutes or 12,096 seconds alongside with corresponding annotations. It also comes with Mel spectrogram version of the data, where an image represents a 1-second window of the original audio, resulting in a total of 17,536 spectrographic images. These are stored in matrix form within .npy files. These are the species covered:
The Topic-Based Paragraph Classification in Genocide-Related Court Transcripts (GTC) dataset is the first reference corpus annotated with samples from genocide tribunals in different international criminal courts. It is made up of witness statements about violence experienced. The material consists of 1475 text passages with about 40 to 120 pages per transcript, covering 3 tribunals: the Extraordinary Chambers in the Courts of Cambodia (ECCC) - 438 pages, the International Criminal Tribunal for Rwanda (ICTR) - 566 pages, and the International Criminal Tribunal of the Former Yugoslavia (ICTY) - 416 pages. As no datasets of any kind containing genocide court transcripts have been published nor other forms of pre-structured or annotated text data in this field of research exist, the aim was to address this gap by providing a systematically annotated dataset.
Visual Knowledge Tracing contains images and human response data for a visual classification task on three datasets. Datasets are intended to serve as benchmark for visual knowledge tracing algorithms.
Teaching assistants (TAs) are heavily used in computer science courses as a way to handle high enrollment and still being able to offer students individual tutoring and detailed assessments. This data is the result of a multi-institutional, multi-national perspective of challenges that TAs in computer science face. 180 reflective essays written by TAs from three institutions across Europe were analyzed and coded. The thematic analysis resulted in five main challenges: becoming a professional TA, student-focused challenges, assessment, defining and using best practice and threats to best practice. In addition, these challenges were all identified within the essays from all three institutions, indicating that the identified challenges are not particularly context-dependent. (2021-04-11)
Suite of OpenAI Gym-compatible multi-agent reinforcement learning environment centered around meta-referential games to benchmark for behavioral traits pertaining to symbolic behaviours, as described in Santoro et al., 2021, "Symbolic Behaviours in Artificial Intelligence", with a primary focus on the following behavioural traits:
YouTube-Hands includes 240 videos which are annotated with hand trajectories.
This dataset contains the data used for all statistical comparisons in our ICSV 2022 submission "Assessment of a Cost-Effective Headphone Calibration Procedure for Soundscape Evaluations", summarised in a single .csv file. <br><br> To obtain the data in this dataset, 17 participants were invited to each rate 27 stimuli twice, once with the stimuli calibrated with a head-and-torso simulator ("HATS method") and once with the stimuli calibrated via an open-circuit voltage method ("OCV method"). This resulted in a total of 17272 = 918 data samples, corresponding to the number of rows in the .csv file. <br><br> For more details on the calibration method and listening test procedure, please refer to our manuscript: <br><br> B. Lam, K. Ooi, K.N. Watcharasupat, Y.-T. Lau, Z.-T. Ong, T. Wong, W.-S. Gan, "Assessment of a Cost-Effective Headphone Calibration Procedure for Soundscape Evaluations", in <i>Proceedings of the 28th International Congress on Sound and Vibration</i>, ICSV28, Singapore, 2
We create a new benchmark called the Smart-City CCTV Violence Detection dataset (SCVD). Current datasets for violence detection contain videos recorded from phone cameras which could alter the needed CCTV distributions. Furthermore, this dataset contains weaponized violence class so it could be used by DNNs to learn the distribution of any potential weapons and infer for quicker action to be carried out on such by the authorities. This means that our dataset is tuned to the fact that any handheld object which could be used to harm humans and properties could be regarded as a weapon
The NMED-H dataset contains scalp EEG responses recorded from 48 adults as they heard intact and scrambled versions of full-length vocal works (Hindi pop songs). Sixteen stimuli were included in the experiment: Four songs in four conditions per song. Twelve participants were assigned to each stimulus, and each participant heard their assigned stimuli twice (24 trials total per stimulus). Dense-array EEG was recorded using the Electrical Geodesics, Inc. (EGI) GES 300 platform. Data are published in Matlab format. The dataset contains (1) raw EEG (individual recordings, 97 files), (2) clean EEG (aggregated by stimulus and listen, 32 files), (3) spatially filtered EEG (aggregated by stimulus condition, four files), (4) behavioral responses (grouped by listen, two files), and (5) participant-stimulus assignment file. Items (1) - (3) are compressed in .zip archives (500 MB - 2 GB each); an example file from each archive can be downloaded separately. Items (4) and (5) are < 1 KB each. This d
We capture some hyperspectral images in our lab using the multishot DD-CASSI architecture. The algorithm can be found on GitHub
The two Coiling Spiral is a 2d classification dataset composed of two classes; each spiral corresponds to one class.
Pavementscapes is a large-scale dataset to develop and evaluate methods for pavement damage segmentation. It is comprised of 4,000 images with a resolution of 1024×2048, which have been recorded in the real-world pavement inspection projects with 15 different pavements. A total of 8,680 damage instances are manually labeled with six damage classes at the pixel level.
A dataset of fashion images for social events