Skip to content

Dataset · enalis

Plant disease benchmark for vision-language models

enalis/LeafBenchV2

A multiple-choice VQA test for evaluating plant pathology understanding in VLMs.

What's inside

LeafBench 2.0 is a visual question answering (VQA) benchmark created from the LeafNet 2.0 dataset. It contains image-question pairs with multiple-choice options covering 9 phytopathology tasks for deep evaluation of plant disease understanding. Already used to test 16 models, including generative VLMs and CLIP-based architectures.

What you can do with it

  • Evaluating the accuracy of generative VLMs and CLIP models in agricultural tasks
  • Comparing multimodal models on their ability to distinguish subtle visual disease features
  • Validating specialized AI assistants for agriculture
Size
10K–100K rows
Downloads this month
38.3K
Likes
4
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("enalis/LeafBenchV2")