The Alzheimer's Disease Multiclass Dataset contains approximately 44,000 MRI images categorized into four distinct classes based on the severity of Alzheimer's disease. This dataset is intended for use in machine learning model training and testing. All images are skull-stripped and clean of non-brain tissue.

Dataset Structure The dataset is organized into the following four directories, each representing a different class of disease severity: NonDemented: Contains 12,800 MRI images of subjects with no signs of dementia. VeryMildDemented: Contains 11,200 MRI images of subjects with very mild symptoms of dementia. MildDemented: Contains 10,000 MRI images of subjects with mild dementia. ModerateDemented: Contains 10,000 MRI images of subjects with moderate dementia.

Image Details Total Number of Images: 44,000 Image Format: MRI scans as .JPG files Image Usage: Suitable for training and testing machine learning models focused on classifying Alzheimer's disease stages.

Disease Severity Classification The dataset follows a severity ranking system for Alzheimer's disease: NonDemented: No dementia. Very Mild Demented: Early signs of dementia, very mild symptoms. Mild Demented: Clear signs of dementia, but still mild. Moderate Demented: More pronounced symptoms of dementia, moderate severity.

This dataset is an augmented and upsampled version of the dataset below: https://www.kaggle.com/datasets/uraninjo/augmented-alzheimer-mri-dataset-v2

This dataset was upsampled as the original dataset had a large class imbalance.

COCO-Counterfactuals is a high quality synthetic dataset for multimodal vision-language model evaluation and for training data augmentation. Each COCO-Counterfactuals example includes a pair of image-text pairs; one is a counterfactual variation of the other. The two captions are identical to each other except a noun subject. The two corresponding synthetic images differ only in terms of the altered subject in the two captions. In our accompanying paper, we showed that the COCO-Counterfactuals dataset is challenging for existing pre-trained multimodal models and significantly increase the difficulty of the zero-shot image-text retrieval and image-text matching tasks. Our experiments also demonstrate that augmenting training data with COCO-Counterfactuals improves OOD generalization on multiple downstream tasks.

Comprehensive Waste Classification Dataset: Hazardous, Non-Recyclable, Organic & Recyclable

Dataset Overview:

This dataset is designed for developing and training computer vision models to recognize and classify various types of waste, with the goal of enhancing environmental sustainability. The dataset contains 30,000+ labeled images across four primary waste categories: Hazardous, Non-Recyclable, Organic, and Recyclable. Each category includes specific subcategories for more precise classification, making it a valuable resource for machine learning and deep learning projects.

How to Use:

For a detailed guide on how to implement and use the dataset in machine learning projects, please visit the project's GitHub repository:
GitHub Repository - CV-based Waste Identifier

Potential Use Cases:

• Training models for automated waste classification
• Research on waste management and environmental sustainability
• Development of smart waste segregation systems
• Educational purposes in environmental studies and machine learning

File and Directory Information:

The dataset is organized in a directory structure where each waste category (and its subcategories) has its own folder. Below is the breakdown of the file organization:

Directory Structure:

• Hazardous: Contains images of dangerous waste materials such as batteries, chemical waste, and medical waste.
• Non-Recyclable: Includes images of materials that cannot be recycled like plastic wrappers, styrofoam, and food cups.
• Organic: Contains images of biodegradable waste, including food scraps and green waste like leaves and garden clippings.
• Recyclable: Includes recyclable waste materials such as paper, glass, and plastic bottles.

Each subcategory folder contains hundreds of images labeled according to the waste type.

Preferred File Formats:

The images in this dataset are provided in common formats, ensuring compatibility with most machine learning frameworks and computer vision models.

Image Formats:

• PNG (.png): Preferred for high-quality images with transparency support.
• JPEG (.jpg, .jpeg): Widely used for compressed, high-quality images with relatively smaller file sizes.
• JPG (.jpg): A commonly used format for standard images, especially for those with no transparency.

These formats are widely supported in most machine learning libraries, including TensorFlow, PyTorch, and OpenCV, and are efficient for model training.

The authors constructed a new UAVDT Dataset focused on complex scenarios with new level challenges. Selected from 10 hours raw videos, about 80, 000 representative frames are fully annotated with bounding boxes as well as up to 14 kinds of attributes (e.g., weather condition, flying altitude, camera view, vehicle category, and occlusion) for three fundamental computer vision tasks: object detection, single object tracking, and multiple object tracking.

ABO

## Overview

ABO is a dataset for object detection tasks - it contains Per Documentation annotations for 8,377 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Trash Filter

## Overview

Trash Filter is a dataset for object detection tasks - it contains Trash annotations for 2,270 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Strawberry LeafDisease

## Overview

Strawberry LeafDisease is a dataset for classification tasks - it contains Strawberry LeafDisease annotations for 1,950 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Viet24

## Overview

Viet24 is a dataset for object detection tasks - it contains Silo Red_ball Blue_ball annotations for 2,494 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Copernicus Dataset

A curated, multi-domain dataset for tokenizer training and small language-model pretraining, covering natural language, source code, and mathematics. The dataset was created as the training corpus for the Copernicus Tokenizer, with the goal of providing broad token and pattern coverage rather than relying on a single text domain.

  Dataset Overview

Property Details

Total size ~12.3 GB

Domains NLP, Code, Mathematics

Format Parquet… See the full description on the dataset page: https://huggingface.co/datasets/Nj-1111/Copernicus-Dataset.

Intrusion Detection Systems (IDS) for IoT and edge environments require datasets with unambiguous labels. Existing corpora often mix benign and malicious traffic within the same capture window, producing ambiguous flow labels that distort model evaluation. This work introduces the TRUST Lab Dataset, a flow-based traffic corpus designed under a single-class session policy: each capture contains exclusively benign traffic or a single attack family, preventing temporal overlap and ensuring label integrity at the bi-flow level. The dataset was generated in an operational testbed reproducing enterprise-grade services (HTTP/S, DNS, email, SSH, SNMP, NTP, MySQL) and modern APIs (REST, GraphQL, SOAP), combined with 15 attack families spanning DDoS/DoS, port scanning, brute force, web/API exploits, DNS abuses, MITM, NIDS evasion, tunneling/exfiltration, C2/beaconing, TLS/SSL anomalies, buffer overflow, and slowloris. Traffic was processed with CICFlowMeter into 16 single-class CSV files totaling ~4.6million bi-flows with 80 features per flow. Comprehensive statistical analyses (distributions, correlation, ANOVA, PCA) confirm discriminative signal without payload inspection. A binary meta-classifier achieves ROC-AUC 0.9676 and recall 0.95, validating TRUST Lab's utility for lightweight edge-oriented IDS evaluation.

License Plate Recognition Dataset

Dataset comprises 83,300+ images of vehicle license plates, meticulously annotated for OCR tasks, object detection, and real-time traffic analysis. Designed to advance research in autonomous vehicles, traffic management, and urban mobility. By leveraging this dataset, researchers and developers can enhance license plate recognition (LPR) systems, improve traffic monitoring, and develop intelligent transportation solutions - Get the data The… See the full description on the dataset page: https://huggingface.co/datasets/UniDataPro/united-kingdom-license-plate-dataset.

Dataset

This dataset was created by Sreafqfqr

Released under CC0: Public Domain

Contents

Мета-датасет: эффективность фенотипов

Сравнительный анализ поведенческих типов агентов. Какой фенотип производит больше артефактов, лучше размножается и эффективнее расходует энергию.

This dataset contains the digitized treatments in Plazi based on the original journal article Carvalho, Pedro Hollanda, Lima, Flávio C. T., Zawadzki, Cláudio H. (2010): Two new species of the Hypostomus cochliodon group (Siluriformes: Loricariidae) from the rio Negro basin in Brazil. Neotropical Ichthyology 8 (1): 39-48, DOI: 10.1590/S1679-62252010000100006, URL: http://www.scielo.br/scielo.php?script=sci_arttext&pid=S1679-62252010000100006&lng=en&tlng=en

Two new species of Hypostomus are described from the middle and upper rio Negro in Brazil. They are assigned to the Hypostomus cochliodon group (sensu Armbruster, 2003) by possessing few spoon-shaped teeth, and dentary angle averaging less than 80°. Hypostomus kopeyaka is described from the rio Tiquié, a tributary of the rio Uaupés, upper rio Negro basin, presents a unique color pattern among the Hypostomus species belonging to the Hypostomus cochliodon group, consisting of conspicuously horizontally elongated, closely-set black spots over the entire dorsal and lateral surfaces of the body. Hypostomus weberi is described from the middle rio Negro and can be distinguished from all remaining Hypostomus species belonging to the Hypostomus cochliodon group by possessing a unique color pattern consisting in large, rounded, widelyspaced black spots over body and fins.

Duas novas espécies de Hypostomus são descritas para a bacia do alto e médio rio Negro no Brasil. Elas são atribuídas ao grupo Hypostomus cochliodon (sensu Armbruster, 2003) por possuírem poucos dentes, em forma de colher e ângulo entre os dentários menor que 80°. Hypostomus kopeyaka, descrita do rio Tiquié, um afluente do rio Uaupés, bacia do alto rio Negro, apresenta um padrão de colorido único entre as espécies de Hypostomus pertencentes ao grupo Hypostomus cochliodon, que consiste em manchas escuras conspícuas horizontalmente alongadas e próximas entre si sobre toda a superfície dorso-lateral do corpo. Hypostomus weberi, descrita do médio rio Negro, é distinguida de todas as espécies de Hypostomus pertencentes ao grupo Hypostomus cochliodon por possuir um padrão de colorido único que consiste em grandes manchas escuras e arredondadas, relativamente afastadas entre si, sobre toda a superfície do corpo e nadadeiras.

BiCE-HG (A bi-conditional egocentric hand gesture dataset under controlled environmental conditions) is a first-person view (FPV) RGB hand gesture dataset collected for augmented reality (AR) and virtual reality (VR) interaction research. The dataset captures 19 distinct hand gestures — 15 single-hand and 4 bimanual — performed by 23 participants across six environmental conditions formed by the factorial combination of two lighting levels (full lighting and low lighting) and three motion scenarios (sitting, standing, and walking).Data were recorded at 3840 × 2160 resolution at 30 FPS using a head-mounted camera providing an egocentric perspective. Each gesture recording provides: raw RGB frames, MediaPipe-annotated frames with hand skeleton overlay, per-frame 3D hand keypoints (21 landmarks × up to 2 hands, normalised x/y/z coordinates), a single representative static frame with corresponding keypoints, and raw and annotated video clipsHand keypoints were extracted using a multi-stage MediaPipe Hands pipeline with additional image preprocessing to improve robustness under challenging lighting and motion conditions. The dataset contains 123,175 frames across 2,607 recordings, achieving an overall frame-level detection rate of 98.44%. Detection rates across the six conditions range from 95.69% (low lighting, standing) to 99.41% (full lighting, walking), demonstrating robustness across all experimental conditions.The 19 gesture classes are organised into 12 action categories covering touch and selection (pinch, air tap), hand shapes (open palm, fist front, fist back), navigation (horizontal swipe, vertical swipe), rotation (wrist rotate), symbolic gestures (thumbs up, OK sign), pointing, greeting (wave, V sign), drawing (clockwise and counter-clockwise circles), zoom (zoom in, zoom out), and framing (hand frame left up, hand frame right up).The dataset is provided in two versions: a dynamic version (keypoint sequences for temporal gesture recognition) and a static version (one representative frame per gesture per recording for pose-based classification). Both versions include 3D hand keypoint files in plain-text format and are accompanied by Python data loaders compatible with PyTorch. An 80/20 train/test split stratified by gesture is recommended for reproducible evaluation.All participants provided written informed consent. Ethics approval was granted by the University of Sheffield, School of Electrical and Electronic Engineering (Reference: 067003, approved 08/07/2025)

VLM-3R Training Data

Training QA data for VLM-3R: vsibench_train/ (VSI-Bench-style tasks) and vstibench_train/ (VSTI-Bench tasks over ScanNet train split).

  Erratum (2026-07-13): corrected camera-position ground truth

A bug in the QA generation pipeline (reported by Jacob Yeung, CMU) extracted the camera center from camera-to-world poses using -R.T @ t instead of pose[:3, 3]. Answers in five vstibench_train files depended on the camera's world position and have… See the full description on the dataset page: https://huggingface.co/datasets/Journey9ni/VLM-3R-DATA.

*********************************

Disclaimer: PLEASE DO NOT USE THIS VERSION (Data has been made unavailable)

Another version, version 2.0.0, of this data set is available here, which contains the original data (+ details on the data set). The version published on the current page contains EEG data that has been pre-processed with a data-driven artifact removal procedure on a trial-by-trial basis. This can introduce slight differences across trials that can possibly be exploited by classifiers. Therefore, for proper benchmarking of auditory attention decoding algorithms that can learn to use such cues, it is recommended to NOT use this pre-processed data set, and instead use the unprocessed version 1.1.0 (the original data).

*********************************

This version of the data set was used in

Das, N., Vanthornhout, J., Francart, T., & Bertrand, A. “Stimulus-aware spatial filtering for single-trial neural response and temporal response function estimation in high-density EEG with applications in auditory research.” NeuroImage, Volume 204, 116211, Jan. 2020
https://doi.org/10.1016/j.neuroimage.2019.116211

The results and conclusions in the paper are not affected by the trial-by-trial application of the artifact removal procedure.

Ambient Scribe Data Set

Ambient Scribe Dataset to Train Multilingual Healthcare AI Models

What’s included: - 199 physician-patient conversations. - Audio, transcription, and summary for each encounter. - Languages: English (UK), Spanish, French, German, Dutch. - Specialties: General Medicine, Psychiatry, Cardiology, OBGYN, Oncology, and more. - Visit types: Routine checkups, new patients, follow-ups, referrals, and medication reviews/

Applications - Training AmbientScribe model. - Healthcare speech-to-text and summarization tools. - Fine-tuning clinical nlp and llms. - Multilingual EHR document assistants. - Prototype agentic workflows for coding. - Benchmark AmbientScribe model accuracy specialists.

The Building TimeSeries (BTS) dataset covers three buildings over a three-year period, comprising more than ten thousand timeseries data points with hundreds of unique Brick classes. Moreover, the metadata is standardized using the Brick schema.To get started, download the data and run the DIEF_inspect_raw.ipynb file.For more info, including data cards: https://github.com/cruiseresearchgroup/DIEF_BTS

Project

## Overview

Project is a dataset for object detection tasks - it contains Door annotations for 954 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [Public Domain license](https://creativecommons.org/licenses/Public Domain).