AgriVerse-Vision

AgriVerse Vision

AgriVerse-Vision is a vision model fine-tuned from google/vit-base-patch16-224 to classify crop health and diagnose 60 plant diseases across 16 crop species.

The model is the primary visual diagnostic component of the AgriVerse agricultural decision-support system.

Performance Highlights

• Test Accuracy: 98.91%
• Test F1-Score (Weighted): 98.90%
• Macro Average Precision / Recall / F1: 0.99 / 0.99 / 0.99
• Test Loss: 0.0483
• Evaluated on: 13,700 held-out test images

Model Overview

• Model Architecture: Vision Transformer (ViT-Base)
• Checkpoint Base: google/vit-base-patch16-224
• Number of Parameters: ~86M
• Number of Classes: 60
• Input Resolution: 224 x 224 pixels (RGB)

Usage

• Foliar plant disease diagnosis and crop health classification
• Field scouting support and agricultural research
• Visual feature extractor for multimodal agricultural systems

Quickstart

Using transformers Pipeline

from PIL import Image
from transformers import pipeline

classifier = pipeline(
    task="image-classification",
    model="Armaan457/AgriVerse-Vision"
)

image = Image.open("leaf_image.jpg").convert("RGB")

predictions = classifier(image, top_k=5)
for pred in predictions:
    print(f"{pred['label']:<32} Confidence: {pred['score']:.4f}")