Alzheimer's Disease Neuroimaging Initiative (ADNI) is a multisite study that aims to improve clinical trials for the prevention and treatment of Alzheimer’s disease (AD).[1] This cooperative study combines expertise and funding from the private and public sector to study subjects with AD, as well as those who may develop AD and controls with no signs of cognitive impairment.[2] Researchers at 63 sites in the US and Canada track the progression of AD in the human brain with neuroimaging, biochemical, and genetic biological markers.[2][3] This knowledge helps to find better clinical trials for the prevention and treatment of AD. ADNI has made a global impact,[4] firstly by developing a set of standardized protocols to allow the comparison of results from multiple centers,[4] and secondly by its data-sharing policy which makes available all at the data without embargo to qualified researchers worldwide.[5] To date, over 1000 scientific publications have used ADNI data.[6] A number of other initiatives related to AD and other diseases have been designed and implemented using ADNI as a model.[4] ADNI has been running since 2004 and is currently funded until 2021.[7] Source: Wikipedia, https://en.wikipedia.org/wiki/Alzheimer%27s_Disease_Neuroimaging_Initiative

Street Scene is a dataset for video anomaly detection. Street Scene consists of 46 training and 35 testing high resolution 1280×720 video sequences taken from a USB camera overlooking a scene of a two-lane street with bike lanes and pedestrian sidewalks during daytime. The dataset is challenging because of the variety of activity taking place such as cars driving, turning, stopping and parking; pedestrians walking, jogging and pushing strollers; and bikers riding in bike lanes. In addition the videos contain changing shadows, moving background such as a flag and trees blowing in the wind, and occlusions caused by trees and large vehicles. There are a total of 56,847 frames for training and 146,410 frames for testing, extracted from the original videos at 15 frames per second. The dataset contains a total of 205 naturally occurring anomalous events ranging from illegal activities such as jaywalking and illegal U-turns to simply those that do not occur in the training set such as pets being walked and a metermaid ticketing a car. Source: A Survey of Single-SceneVideo Anomaly Detection Image Source: https://www.merl.com/demos/video-anomaly-detection

Real 3D-AD is the first point cloud anomaly detection dataset for industrial products. Real3D-AD comprises a total of 1,254 samples that are distributed across 12 distinct categories. These categories include Airplane, Car, Candybar, Chicken, Diamond, Duck, Fish, Gemstone, Seahorse, Shell, Starfish, and Toffees. Each training sample is an absence of blind spots, and a realistic, high-accuracy prototype.

Encourages machine learning research in this area and to help facilitate further work in understanding and mitigating the effects of climate change. Source: ExtremeWeather: A large-scale climate dataset for semi-supervised detection, localization, and understanding of extreme weather events

Includes 5,824 fundus images labeled with either positive glaucoma (2,392) or negative glaucoma (3,432). Source: Attention Based Glaucoma Detection: A Large-scale Database and CNN Model

Click to add a brief description of the dataset (Markdown and LaTeX enabled). Provide: * a high-level explanation of the dataset characteristics * explain motivations and summary of its content * potential use cases of the dataset

The NINCO (No ImageNet Class Objects) dataset is introduced in the ICML 2023 paper In or Out? Fixing ImageNet Out-of-Distribution Detection Evaluation. The images in this dataset are free from objects that belong to any of the 1000 classes of ImageNet-1K (ILSVRC2012), which makes NINCO suitable for evaluating out-of-distribution detection on ImageNet-1K . The NINCO main dataset consists of 64 OOD classes with a total of 5879 samples. These OOD classes were selected to have no categorical overlap with any classes of ImageNet-1K. Each sample was inspected individually by the authors to not contain ID objects. Besides NINCO, included are (in the same .tar.gz file) truly OOD versions of 11 popular OOD datasets with in total 2715 OOD samples. Further included are 17 OOD unit-tests, with 400 samples each. Code for loading and evaluating on each of the three datasets is provided at https://github.com/j-cb/NINCO. When using NINCO, please consider citing (besides the bibtex given below) the following data sources that were used to create NINCO: Hendrycks et al.: ”Scaling out-of-distribution detection for real-world settings”, ICML, 2022. Bossard et al.: ”Food-101 – mining discriminative components with random forests”, ECCV 2014. Zhou et al.: ”Places: A 10 million image database for scene recognition”, IEEE PAMI 2017. Huang et al.: ”Mos: Towards scaling out-of-distribution detection for large semantic space”, CVPR 2021. Li et al.: ”Caltech 101 (1.0)”, 2022. Ismail et al.: ”MYNursingHome: A fully-labelled image dataset for indoor object classification.”, Data in Brief (V. 32) 2020. The iNaturalist project: https://www.inaturalist.org/ When using NINCO_popular_datasets_subsamples, additionally to the above, please consider citing: Cimpoi et al.: ”Describing textures in the wild”, CVPR 2014. Hendrycks et al.: ”Natural adversarial examples”, CVPR 2021. Wang et al.: ”Vim: Out-of-distribution with virtual-logit matching”, CVPR 2022. Bendale et al.: ”Towards Open Set Deep Networks”, CVPR 2016. Vaze et al.: ”Open-set Recognition: a Good Closed-set Classifier is All You Need?”, ICLR 2022. Wang et al.: ”Partial and Asymmetric Contrastive Learning for Out-of-Distribution Detection in Long-Tailed Recognition.” ICML, 2022. Galil et al.: “A framework for benchmarking Class-out-of-distribution detection and its application to ImageNet”, ICLR 2023.

The GoodsAD dataset contains 6124 images with 6 categories of common supermarket goods. Each category contains multiple goods. All images are acquired with 3000 × 3000 high-resolution. The object locations in the images are not aligned. Most objects are in the center of the images and one image only contains a single object. Most anomalies occupy only a small fraction of image pixels. Both image-level and pixel-level annotations are provided. Each image is named with 6 digits, with the first three digits representing the category of the product and the last three representing the serial number. The dataset format is same as MVTec AD.

The dataset is constructed from images of defective production items that were provided and annotated by Kolektor Group d.o.o.. The images were captured in a controlled industrial environment in a real-world case. The dataset consists of 399 images at 500 x ~1250 px in size. Please cite our paper published in the Journal of Intelligent Manufacturing when using this dataset: @article{Tabernik2019JIM, author = {Tabernik, Domen and {\v{S}}ela, Samo and Skvar{\v{c}}, Jure and Sko{\v{c}}aj, Danijel}, journal = {Journal of Intelligent Manufacturing}, title = {{Segmentation-Based Deep-Learning Approach for Surface-Defect Detection}}, year = {2019}, month = {May}, day = {15}, issn={1572-8145}, doi={10.1007/s10845-019-01476-x} }

KolektorSDD2 is a surface-defect detection dataset with over 3000 images containing several types of defects, obtained while addressing a real-world industrial problem. The dataset consists of: * 356 images with visible defects * 2979 images without any defect * image sizes of approximately 230 x 630 pixels * train set with 246 positive and 2085 negative images * test set with 110 positive and 894 negative images * several different types of defects (scratches, minor spots, surface imperfections, etc.)

RoadAnomaly21 is a dataset for anomaly segmentation, the task of identify the image regions containing objects that have never been seen during training. It consists of an evaluation dataset of 100 images with pixel-level annotations. Each image contains at least one anomalous object, e.g. animals or unknown vehicles. The anomalies can appear anywhere in the image and widely differ in size, covering from 0.5% to 40% of the image

ToyADMOS dataset is a machine operating sounds dataset of approximately 540 hours of normal machine operating sounds and over 12,000 samples of anomalous sounds collected with four microphones at a 48kHz sampling rate, prepared by Yuma Koizumi and members in NTT Media Intelligence Laboratories. The ToyADMOS dataset is designed for anomaly detection in machine operating sounds (ADMOS) research. It is designed for three tasks of ADMOS: product inspection (toy car), fault diagnosis for fixed machine (toy conveyor), and fault diagnosis for moving machine (toy train). Source: https://github.com/YumaKoizumi/ToyADMOS-dataset

The dataset contains 2,624 samples of $300\times300$ pixels 8-bit grayscale images of functional and defective solar cells with varying degree of degradations extracted from 44 different solar modules. The defects in the annotated images are either of intrinsic or extrinsic type and are known to reduce the power efficiency of solar modules. All images are normalized with respect to size and perspective. Additionally, any distortion induced by the camera lens used to capture the EL images was eliminated prior to solar cell extraction.

The human-Related version of the ShanghaiTech Campus, was first presented by Morais et al. in the paper "Learning Regularity in Skeleton Trajectories for Anomaly Detection in Videos".

The UCR Anomaly Archive is a collection of 250 uni-variate time series collected in human medicine, biology, meteorology and industry. The collected time series contain a few natural anomalies though the majority of the anomalies are artificial . The dataset was first used in an anomaly detection contest preceding the ACM SIGKDD conference 2021. Each of the time series contains exactly one, occasionally subtle anomaly after a given time stamp. The data before that timestamp can be considered normal. The time series collected in the UCR Anomaly Archive can be categorized into 12 types originating from the four domains human medicine, meteorology, biology and industry. The distribution across the domains is highly imbalanced with around 64% of the times series being collected in human medicine applications, 22% in biology, 9% in industry and 5% being air temperature measurements. The time series within a single type (e.g. ECG) are not completely unique, but differ in terms of injected anomalies or a modification of the original time series through added Gaussian noise and wandering baselines. The downloadable archive contains, among other supplemental material, a set of slides explaining the injected anomalies with examples.

For benchmarking, please refer to its variant UPFD-POL and UPFD-GOS. The dataset has been integrated with Pytorch Geometric (PyG) and Deep Graph Library (DGL). You can load the dataset after installing the latest versions of PyG or DGL. The UPFD dataset includes two sets of tree-structured graphs curated for evaluating binary graph classification, graph anomaly detection, and fake/real news detection tasks. The dataset is dumped in the form of Pytorch-Geometric dataset object. You can easily load the data and run various GNN models using PyG. The dataset includes fake&real news propagation (retweet) networks on Twitter built according to fact-check information from Politifact and Gossipcop. The news retweet graphs were originally extracted by FakeNewsNet. Each graph is a hierarchical tree-structured graph where the root node represents the news; the leaf nodes are Twitter users who retweeted the root news. A user node has an edge to the news node if he/she retweeted the news tweet. Two user nodes have an edge if one user retweeted the news tweet from the other user. We crawled near 20 million historical tweets from users who participated in fake news propagation in FakeNewsNet to generate node features in the dataset. We incorporate four node feature types in the dataset, the 768-dimensional bert and 300-dimensional spacy features are encoded using pretrained BERT and spaCy word2vec, respectively. The 10-dimensional profile feature is obtained from a Twitter account's profile. You can refer to profile_feature.py for profile feature extraction. The 310-dimensional content feature is composed of a 300-dimensional user comment word2vec (spaCy) embedding plus a 10-dimensional profile feature. The dataset statistics is shown below: | Data | #Graphs | #Fake News| #Total Nodes | #Total Edges | #Avg. Nodes per Graph | |-------|--------|--------|--------|--------|--------| | Politifact | 314 | 157 | 41,054 | 40,740 | 131 | | Gossipcop | 5464 | 2732 | 314,262 | 308,798 | 58 | Please refer to the paper for more details about the UPFD dataset. Due to the Twitter policy, we could not release the crawled user's historical tweets publicly. To get the corresponding Twitter user information, you can refer to the news lists under \data in our github repo and map the news id to FakeNewsNet. Then, you can crawl the user information by following the instruction on FakeNewsNet. In the UPFD project, we use Tweepy and Twitter Developer API to get the user information.

Yelp-Fraud is a multi-relational graph dataset built upon the Yelp spam review dataset, which can be used in evaluating graph-based node classification, fraud detection, and anomaly detection models. - Dataset Statistics | # Nodes | %Fraud Nodes (Class=1) | |-------|--------| | 45,954 | 14.5 | | Relation | # Edges | |--------|--------| | R-U-R | 49,315 | | R-T-R | 573,616 | | R-S-R | 3,402,743 | | All | 3,846,979 | - Graph Construction The Yelp spam review dataset includes hotel and restaurant reviews filtered (spam) and recommended (legitimate) by Yelp. We conduct a spam review detection task on the Yelp-Fraud dataset which is a binary classification task. We take 32 handcrafted features from SpEagle paper as the raw node features for Yelp-Fraud. Based on previous studies which show that opinion fraudsters have connections in user, product, review text, and time, we take reviews as nodes in the graph and design three relations: 1) R-U-R: it connects reviews posted by the same user; 2) R-S-R: it connects reviews under the same product with the same star rating (1-5 stars); 3) R-T-R: it connects two reviews under the same product posted in the same month. To download the dataset, please visit this Github repo. For any other questions, please email ytongdou(AT)gmail.com for inquiry.

HyperKvasir dataset contains 110,079 images and 374 videos where it captures anatomical landmarks and pathological and normal findings. A total of around 1 million images and video frames altogether.

Unlike previous datasets that focus on detecting the diversity of defect categories (like MVTec AD and VisA), AeBAD is centered on the diversity of domains within the same data category. The aim of AeBAD is to automatically detect abnormalities in the blades of aero-engines, ensuring their stable operation. AeBAD consists of two sub-datasets: the single-blade dataset (AeBAD-S) and the video anomaly detection of blades (AeBAD-V). AeBAD-S comprises images of single blades of different scales, with a primary feature being that the samples are not aligned. Furthermore, there is a domain shift between the distribution of normal samples in the test set and the training set, where the domain shifts are mainly caused by the changes in illumination and view. AeBAD-V, on the other hand, includes videos of blades assembled on the blisks of aero-engines, with the aim of detecting blade anomalies during blisk rotation. A distinctive feature of AeBAD-V is that the shooting view in the test set differs from that in the training set.

Casia V1 is a dataset for forgery classification. Casia V1+ is a modification of the Casia V1 dataset proposed by Chen et al. that replaces authentic images that also exist in Casiav2 with images from the COREL dataset to avoid data contamination.

A dataset consisting of stereo thermal, stereo color, and cross-modality image pairs with high accuracy ground truth (< 2mm) generated from a LiDAR. The authors scanned 100 cluttered indoor and 80 outdoor scenes featuring challenging environments and conditions. CATS contains approximately 1400 images of pedestrians, vehicles, electronics, and other thermally interesting objects in different environmental conditions, including nighttime, daytime, and foggy scenes. Source: CATS: A Color and Thermal Stereo Benchmark

Darpa is a dataset consisting of communications between source IPs and destination IPs. This dataset contains different attacks between IPs. Source: dynnode2vec: Scalable Dynamic Network Embedding Image Source: https://archive.ll.mit.edu/ideval/files/1999_DARPA_EvaulationSumPlans.pdf

Predicting forest cover type from cartographic variables only (no remotely sensed data). The actual forest cover type for a given observation (30 x 30 meter cell) was determined from US Forest Service (USFS) Region 2 Resource Information System (RIS) data. Independent variables were derived from data originally obtained from US Geological Survey (USGS) and USFS data. Data is in raw form (not scaled) and contains binary (0 or 1) columns of data for qualitative independent variables (wilderness areas and soil types). This study area includes four wilderness areas located in the Roosevelt National Forest of northern Colorado. These areas represent forests with minimal human-caused disturbances, so that existing forest cover types are more a result of ecological processes rather than forest management practices. Some background information for these four wilderness areas: Neota (area 2) probably has the highest mean elevational value of the 4 wilderness areas. Rawah (area 1) and Comanche Peak (area 3) would have a lower mean elevational value, while Cache la Poudre (area 4) would have the lowest mean elevational value. As for primary major tree species in these areas, Neota would have spruce/fir (type 1), while Rawah and Comanche Peak would probably have lodgepole pine (type 2) as their primary species, followed by spruce/fir and aspen (type 5). Cache la Poudre would tend to have Ponderosa pine (type 3), Douglas-fir (type 6), and cottonwood/willow (type 4). The Rawah and Comanche Peak areas would tend to be more typical of the overall dataset than either the Neota or Cache la Poudre, due to their assortment of tree species and range of predictive variable values (elevation, etc.) Cache la Poudre would probably be more unique than the others, due to its relatively low elevation range and species composition.

The human-Related version of the CUHK Avenue dataset, first presented by Morais et al. in the paper "Learning Regularity in Skeleton Trajectories for Anomaly Detection in Videos".

a dataset of time-series anomaly detection

Five datasets used in NeurTraL-AD paper: \textit{RacketSports (RS).} Accelerometer and gyroscope recording of players playing four different racket sports. Each sport is designated as a different class. \textit{Epilepsy (EPSY).} Accelerometer recording of healthy actors simulating four different activity classes, one of them being an epileptic shock. \textit{Naval air training and operating procedures standardization (NAT).} Positions of sensors mounted on different body parts of a person performing activities. There are six different activity classes in the dataset. \textit{Character trajectories (CT).} Velocity trajectories of a pen on a WACOM tablet. There are $20$ different characters in this dataset. \textit{Spoken Arabic Digits (SAD).} MFCC features of ten arabic digits spoken by $88$ different speakers.

COVERAGE contains copymove forged (CMFD) images and their originals with similar but genuine objects (SGOs). COVERAGE is designed to highlight and address tamper detection ambiguity of popular methods, caused by self-similarity within natural images. In COVERAGE, forged–original pairs are annotated with (i) the duplicated and forged region masks, and (ii) the tampering factor/similarity metric. For benchmarking, forgery quality is evaluated using (i) computer vision-based methods, and (ii) human detection performance. Source: COVERAGE

Contains normal driving videos together with a set of anomalous actions in its training set. In the test set of the DAD dataset, there are unseen anomalous actions that still need to be winnowed out from normal driving. Source: Driver Anomaly Detection: A Dataset and Contrastive Learning Approach

Contains 4,677 videos with temporal, spatial, and categorical annotations. Source: When, Where, and What? A New Dataset for Anomaly Detection in Driving Videos

The Standardized Project Gutenberg Corpus (SPGC) is an open science approach to a curated version of the complete PG data containing more than 50,000 books and more than 3×109 word-tokens. Source: https://arxiv.org/abs/1812.08092

Amazon-Fraud is a multi-relational graph dataset built upon the Amazon review dataset, which can be used in evaluating graph-based node classification, fraud detection, and anomaly detection models. - Dataset Statistics | # Nodes | %Fraud Nodes (Class=1)| |-------|--------| | 11,944 | 9.5 | | Relation | # Edges | |--------|--------| | U-P-U | 175,608 | | U-S-U | 3,566,479 | | U-V-U | 1,036,737 | | All | 4,398,392 | - Graph Construction The Amazon dataset includes product reviews under the Musical Instruments category. Similar to this paper, we label users with more than 80% helpful votes as benign entities and users with less than 20% helpful votes as fraudulent entities. we conduct a fraudulent user detection task on the Amazon-Fraud dataset, which is a binary classification task. We take 25 handcrafted features from this paper as the raw node features for Amazon-Fraud. We take users as nodes in the graph and design three relations: 1) U-P-U: it connects users reviewing at least one same product; 2) U-S-V: it connects users having at least one same star rating within one week; 3) U-V-U: it connects users with top 5% mutual review text similarities (measured by TF-IDF) among all users. To download the dataset, please visit this Github repo. For any other questions, please email ytongdou(AT)gmail.com for inquiry.

The dataset contains transactions made by credit cards in September 2013 by European cardholders. This dataset presents transactions that occurred in two days, where we have 492 frauds out of 284,807 transactions. The dataset is highly unbalanced, the positive class (frauds) account for 0.172% of all transactions. It contains only numerical input variables which are the result of a PCA transformation. Unfortunately, due to confidentiality issues, we cannot provide the original features and more background information about the data. Features V1, V2, … V28 are the principal components obtained with PCA, the only features which have not been transformed with PCA are 'Time' and 'Amount'. Feature 'Time' contains the seconds elapsed between each transaction and the first transaction in the dataset. The feature 'Amount' is the transaction Amount, this feature can be used for example-dependent cost-sensitive learning. Feature 'Class' is the response variable and it takes value 1 in case of fraud and 0 otherwise. Given the class imbalance ratio, we recommend measuring the accuracy using the Area Under the Precision-Recall Curve (AUPRC). Confusion matrix accuracy is not meaningful for unbalanced classification.

CHAD: Charlotte Anomaly Dataset CHAD is high-resolution, multi-camera dataset for surveillance video anomaly detection. It includes bounding box, Re-ID, and pose annotations, as well as frame-level anomaly labels, dividing all frames into two groups of anomalous or normal. You can find the paper with all the details in the following link: CHAD: Charlotte Anomaly Dataset. Please refer to the page of the dataset for more information.

The Human Related version of UBnormal ("UBnormal: New Benchmark for Supervised Open-Set Video Anomaly Detection," Acsintoae et al.) was introduced by Flaborea et al. in the paper "Contracting Skeletal Kinematics for Human-Related Video Anomaly Detection".

UBI-Fights - Concerning a specific anomaly detection and still providing a wide diversity in fighting scenarios, the UBI-Fights dataset is a unique new large-scale dataset of 80 hours of video fully annotated at the frame level. Consisting of 1000 videos, where 216 videos contain a fight event, and 784 are normal daily life situations. All unnecessary video segments (e.g., video introductions, news, etc.) that could disturb the learning process were removed.

We introduce a new dataset, Watch and Learn Time-lapse (WALT), consisting of multiple (4K and 1080p) cameras capturing urban environments over a year.