Skip to content

Dataset · axxkaya

Descriptive instructions for computer vision tasks

axxkaya/UVT-Terminological-based-Vision-Tasks

Large image-text pair corpus with detailed instructions for training CV models for zero-shot generalization.

What's inside

The dataset contains between 1 and 10 million pairs of images and text instructions in English. Instead of rigid class names, it uses detailed explanatory instructions to teach computer vision models to understand new tasks in zero-shot mode.

What you can do with it

  • Training universal vision-language models capable of solving new CV tasks via text description
  • Researching zero-shot generalization in computer vision tasks
Languages
English
Size
1M–10M rows
Downloads this month
1.4K
Likes
41

How to load it

from datasets import load_dataset

ds = load_dataset("axxkaya/UVT-Terminological-based-Vision-Tasks")