Skip to content

Dataset · nvidia

Nemotron VL training data

nvidia/Nemotron-Image-Training-v3.5

Collection of multimodal training annotations for the Nemotron Super 3.5 VL model.

What's inside

Combines 40 data subsets containing 6,540,914 annotations and 29,991 synthetic English image files. Covers detection, scene text understanding, spatial and graphical analysis, planning, and video question-answering tasks.

What you can do with it

  • Training and fine-tuning multimodal vision-language models (VLMs)
  • Developing visual search and image reasoning systems
  • Creating vision-enabled robotics and planning agents
Tasks
Visual question answering, Image text → text
Languages
English
Size
1M–10M rows
Downloads this month
2.4K
Likes
8

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/Nemotron-Image-Training-v3.5")