Datasets:
metadata
license: cc-by-nc-3.0
language:
- en
task_categories:
- image-to-text
- image-classification
- zero-shot-classification
tags:
- Agriculture
pretty_name: AgroInstruct
AgroInstruct
We develop a diverse expert tuning set (~70K) comprising conversations about different agricultural subjects. Specifically, our pipeline uses vision-only datasets from four domains (diseases, weeds, insects, and fruits). It synthesizes AgroInstruct, which consists of three types of questions: complex multi-turn conversations, simple questions, and image descriptions.
📥 Download the Training Dataset: Access our AgroInstruct dataset.