Chart-to-code generation benchmark for MLLMs
Chart images and English texts to evaluate how multimodal models write code.
- Text generation
- Text → image
- Image → text
- 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
Chart images and English texts to evaluate how multimodal models write code.
Images and questions for cross-domain VQA in industrial electronics defect detection tasks.
Images of medieval manuscripts and their text transcriptions in six European languages.
A large multimodal set of 1 million to 10 million image-text pairs in English.
80K image-instruction pairs with responses from 12 vision-language models evaluated by GPT-4V.
From 1 to 10 million document image-text pairs in English for training VLM models.
Large web dataset for detecting and classifying input fields on document scans and pages.
37.6 million image-text pairs from Wikipedia across 108 languages for multimodal retrieval and training.
Manually annotated video, audio, and images for evaluating deepfake detectors in real-world conditions.
A dataset of 31,500 remote sensing images for 45-class scene classification.
A large open dataset of 218M images with expanded descriptions enriched with web facts.
English images and text for training and validating trustworthy multimodal models via RLAIF.
1,000 interior 3D scenes in USDZ format for simulation and robotics in NVIDIA Isaac Sim.
A set of PC screenshots with interface element markup and GPT-4o descriptions for UI-grounding tasks.
Optimized 3D models of NASA spacecraft, rovers, and spacesuits for engines and the web.
1.8M multimodal examples with detailed reasoning distilled from Qwen3-VL-235B-Thinking.
6k+ top-voted NSFW images from CivitAI with prompt and NSFW-level metadata.
The LongDS-Bench multimodal benchmark (up to 1,000 examples) tests how agents maintain context during long data analysis sessions.
49,662 real mechanical engineering samples in four formats: PDF drawing, PNG image, STEP 3D model, and CadQuery code.
Printed sheet music images with region annotations for testing optical music recognition systems.
A dataset of 1K–10K image and text instruction pairs for testing image editing models on real-world scenarios.
Millions of pairs of images, object descriptions, and point coordinates for training models to point out details.
400,476 videos with previews, view counts, explicit flags, and captions for video-text models.
Monthly collection of BBC News articles and images from 2017 onwards for training and evaluating language and multimodal models.