Billion-scale image-text pair metadata
Metadata and image URLs with text descriptions exceeding one billion records for multimodal learning.
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
Metadata and image URLs with text descriptions exceeding one billion records for multimodal learning.
Benchmark for evaluating spatial reasoning of cross-species models.
Processed training dataset for computer vision tasks with target encoding in RGB images.
Large multimodal dataset with images and vibration data of railway infrastructure.
Classic dataset of over 108k images for classifying 397 scene categories.
Collection of tens of millions of images for training generated image detectors with commercially clean licenses.
Evaluation set of images and English text across 10 tasks and 30 subtasks for testing U-MLLMs.
Millions of photos up to 4K with Creative Commons licenses and synthetic English captions.
Images and knowledge regarding players, teams, referees, and stadiums for soccer analysis.
A benchmark of 46k Bangalore traffic frames with annotations for 14 transport classes for object detection.
Over 120,000 annotated frames from 100 surgeries for anatomical recognition in surgical video.
Images and text in English and Chinese for training and evaluating multimodal models.
Formatted version of the POPE test (10k-100k examples) for evaluating object hallucinations in VLMs.
Archive of over 11,000 past exam questions from Ramdeobaba University for text-retrieval tasks.
A large dataset of image-text pairs without re-compression for training diffusion models.
The 70,000-sample MNIST dataset formatted for use with FiftyOne.
Dataset for benchmarking counterfactual vision-language models (VLMs).
Copy of the well-known TuSimple benchmark image set for computer vision tasks.
A large dataset of aerial urban panoramic images and matching depth maps for 3D scene understanding.
A large collection of high-resolution Yahoo Flickr images with English descriptions for generative models.
About 1.3 TB of raw radio frequency data from 37 different UAVs for drone detection and identification tasks.
A multimodal English science question dataset with multiple-choice options and images.
An image dataset accompanying a NeurIPS 2024 paper for detecting lip-sync deepfakes.
12,780 examples for Visual Question Answering (VQA) combined from multiple distinct sources for convenient model evaluation.