Home › Productivity › New Plant Disease Detection

New Plant Disease Detection - AI Productivity Tool

Overview

New Plant Disease Detection is a Hugging Face Space (author: etahamad) that lets users upload plant leaf images and returns an automatic disease diagnosis, a numeric confidence score, and visual highlights on the leaf that indicate regions used for the prediction. The web-based tool is designed for quick, per-image checks and is pitched toward gardeners, agronomists, students, and researchers who need a fast, visual second opinion for symptomatic leaves. The Space’s public page shows minimal metadata: as of this listing it has 4 likes and 0 downloads on Hugging Face, but the Space does not publish model architecture, training data, or evaluation metrics on the public page. The interface focuses on image input and visual output (saliency/heatmap-style highlights) rather than exposing model internals, making it convenient for non-technical users but limited for users seeking reproducible model details.

Model Statistics

Likes: 4

Model Details

The Hugging Face Space does not publish explicit technical details for the underlying model (architecture, parameter count, or training dataset) on the public page. The Space author (etahamad) provides an interactive image upload -> inference UI that returns a predicted disease label, a confidence/probability score, and a visual overlay highlighting important regions of the leaf used for the decision. The page also does not list an exposed pipeline type (classification/detection/segmentation) or any downloadable model artifact. Because the Space itself omits architecture and dataset information, users who require reproducibility or benchmarking should contact the author via the Hugging Face profile or look for a linked GitHub/README if present on the Space. Typical systems that provide the same outputs commonly use convolutional neural networks (e.g., MobileNet, EfficientNet) fine-tuned on plant leaf datasets (for example, PlantVillage), with saliency techniques (Grad-CAM, integrated gradients) to generate visual highlights — but these approaches are reported here as common practice, not as confirmed details for this specific Space.

Key Features

Upload single leaf images through a web UI for per-image diagnosis
Returns a predicted disease label for the uploaded leaf image
Provides a numeric confidence score for each prediction
Generates visual highlights (saliency/heatmap) showing influential regions
Accessible as a Hugging Face Space — no local install required

Example Usage

Example (python):

from gradio_client import Client

# Create a client for the Hugging Face Space
client = Client("etahamad/new-plant-disease-detection")

# Replace with the path to a local leaf image
image_path = "./leaf.jpg"

# Call the Space's default predict function. Inputs vary by Space; this calls the main interface.
# The return format depends on the Space implementation (label, confidence, and visual overlay are typical).
result = client.predict(image_path)

print("Raw result from Space:", result)

# Example post-processing (structure depends on the Space):
# If the space returns a dict with keys ['label','confidence','mask'] you can do:
# label = result.get('label')
# confidence = result.get('confidence')
# mask = result.get('mask')
# print(label, confidence)

# Note: Input names and return structure depend on the Space implementation; adapt as needed.
# If you get an error, open the Space in a browser to confirm the expected inputs and API name.