Multi-image understanding benchmark for VLMs
English visual question answering dataset for evaluating multimodal models' ability to analyze multiple images simultaneously.
- Question answering
- English
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
English visual question answering dataset for evaluating multimodal models' ability to analyze multiple images simultaneously.
A large-scale human motion dataset extracted from web video with detailed descriptions for robotics and animation generation.
SVG illustration examples with text for training and evaluating text-to-SVG and image-to-SVG models.
English dataset for evaluating models in Visual Question Answering and table comprehension tasks.
A set of 326 motion files in GLB format based on the Microsoft Rocketbox avatar library.
595K image-text pairs from CC3M for feature alignment when training multimodal models.
Corpus of image-instruction pairs for SFT training of vision-language models in English and Chinese.
Large set of Flickr image links in public domain and CC0 for training.
Collection of Stable Diffusion images and tabular metadata from Civitai platform for generative models.
A dataset for relational and analogical visual reasoning.
A massive 12 TB web corpus from 2024–2025 for pre-training multimodal models.
Collection of 14 open benchmarks for text recognition and detection in images.
17,706 image and mask pairs for pixel-level segmentation of clothing, accessories, and human body parts.
A collection of open medical images for training and testing computer vision models.
8,000 generated images with prompts and evaluations from eight text-to-image models to compare their artistic qualities.
Images and segmentation masks of medieval manuscripts for locating text lines and blocks using the SegmOnto standard.
Small set of Pokemon images with bilingual captions for training text-to-image models.
A small document QA benchmark across eight domains for multimodal reasoning.
An image-text dataset for testing how accurately MLLMs localize objects.
100k to 1M visually diverse English math problems for training VLM reasoning.
13k images from OpenAI 4o with over 200k real human ratings for training preference models.
Metadata of nearly 660k Suno AI songs with multilingual descriptions and audio modality.
Over 115k illustrated pages from historical editions of the Britannica encyclopedia (1768–1929).
Audio recordings with spectrograms and metadata for classifying rare and traditional instruments of China.