Loading...
Loading...
Image classification, object detection, segmentation, face recognition, OCR, image generation, video understanding
17,653 datasets
IMAGDressing is a dataset and framework for interactive modular apparel generation and virtual try-on, released by muzishen for AAAI 2025. It facilitates high-fidelity human image generation with specific controls for garments, poses, and scenes to ensure consistency in virtual dressing applications.
French metropolitan territory agricultural parcels annotated for panoptic segmentation. The dataset contains 2,433 patches, each representing a Sentinel-2 multispectral image time series of variable length. It was authored by IGNF and last updated on the Hugging Face platform in August 2025.
T2I Imagenet Normal is a dataset hosted on HuggingFace by arijitghosh, last updated on October 2, 2025. The title suggests it relates to text-to-image generation and may contain normal map data derived from or associated with the ImageNet dataset. The specific content, scale, and structure require verification after download as metadata is minimal.
Ambient Diffusion Omni (Ambient-o) is a framework for training diffusion models using low-quality, synthetic, and out-of-distribution images. The dataset, created by author adrianrm, was last updated on August 17, 2025. It is hosted on the Hugging Face platform.
Control-Tax is a dataset containing model-generated backdoor solutions and exploits for AI safety research. The dataset is based on problems from the APPS dataset and was generated for use with ControlArena. It was created by RoganInglis and last updated on August 27, 2025.
Pre-tokenized data for 3D scenes and images, generated via a pipeline described in the repository. The dataset is hosted on Hugging Face by user aadarsh99 and was last updated on August 27, 2025. Specific details on data volume, file formats, and licensing are not provided in the available metadata.
A dataset of images related to computer-aided design (CAD). It was published on the Hugging Face platform by the user turiya-ai and was last updated on October 4, 2025. The specific content, scale, and file formats are not detailed in the available metadata.
The Micro-Action-52 (MA-52) dataset is designed specifically for Micro-Action Recognition research. It is hosted on Hugging Face by the author kunli-cs and was last updated on August 30, 2025. The dataset is intended for non-commercial scientific purposes and requires a formal application for access.
TaskStream is a dataset of enterprise business workflows, decision-making processes, organizational structures, and operational documentation across multiple industries. It was uploaded to Hugging Face by vrushankpatel5 and last updated on September 30, 2025. The specific scale, file formats, and license details are not provided in the available metadata.
A dataset for optical character recognition (OCR) tasks in the Bengali (Bangla) language. It was uploaded by user 'swapnil7777' to the Hugging Face platform and was last updated on October 8, 2025. The specific content, size, and structure of the dataset require verification after download.
EmbodiedOcc-ScanNet is a reorganized benchmark dataset for evaluating embodied 3D occupancy prediction tasks. It is based on local annotations from the ScanNet dataset and accompanies the paper 'EmbodiedOcc: Embodied 3D Occupancy Prediction for Vision-based Online Scene Understanding'. The dataset was created by YkiWu and was last updated on Hugging Face on 2025-08-27.
GPT-Image-Edit-1.5M consists of 1.5 million image editing pairs produced by UCSC-VLAA in 2025. The collection aggregates samples from HQ-Edit, UltraEdit, OmniEdit, and Complex-Edit, with all output images systematically regenerated using the GPT-Image-1 model.
Syamsuddin's dataset contains Indonesian administrative region data (Province, Regency/City, District, Village/Urban Village) extracted from the wilayah_2023.sql file, sourced from the Ministry of Home Affairs via a public repository. The data is formatted specifically for instruction tuning of the Qwen large language model, with fields for instruction, input, and output. It was last updated on August 27, 2025.
VINCIE-10M is a dataset for unlocking in-context image editing from video. It was created by researchers including Leigang Qu and Feng Cheng, using chain-of-thought prompting with a Vision-Language Model to annotate visual transitions between frames. The dataset was last updated on August 24, 2025.
Six publicly available whole-slide image (WSI) datasets are included, with embeddings generated using two different encoders: ResNet-50 and Conch. The repository contains embedding superpixel segmentation and checkpoint data, authored by zeyugao and last updated on September 4,ๆไปฌๅ็ฐ2025. Each dataset is available as a ZIP file containing both the embeddings and the corresponding superpixel segmentation results.
A dataset of line art images depicting athletes in sportswear, created by author randt. The collection is described as containing multiple views of sportsmen and was last updated on August 26, 2025. It is hosted on the Hugging Face platform.
Persian Ocr Synth Sentences 100K is a text dataset published on the HuggingFace platform by WeightedAI and last updated on October 5, 2025. The title suggests it likely contains 100,000 synthetically generated sentences in the Persian (Farsi) language, potentially designed for optical character recognition tasks. Specific details regarding the data's structure, source, and generation methodology are not provided in the available metadata.
22,926,427 instance masks of individual agricultural fields are provided across 672,909 high-resolution satellite image patches. The dataset, created by MykolaL, excludes Pleiades satellite imagery due to third-party policies but maintains broad coverage. It was last updated on August 22, 2025.
Annotated interaction data extracted from video frames for detecting violent vs non-violent human interactions. The dataset, created by harmesh95, was last updated on August 17, 2025. It includes bounding boxes, pose keypoints, motion features, and violence indicators for pairs of interacting persons.
Kinetics-700 is a large-scale collection of YouTube video URLs curated for human action recognition. The dataset was uploaded by atalaydenknalbant and was last updated on the Hugging Face platform in August 2025. It consists of 22 compressed archives that must be downloaded and decompressed to access the complete video collection.