🌻 Hacktoberfest Open-Source AI Challenge Week 1: “Touch Grass”

Healthy Plants Start With Smart Care.

A 100% locally runnable, open-weight plant disease screening companion. Identify leaf diseases with AI, follow cautious organic care guidance, and step outside to reconnect with living nature.

🔬 Try Live Analyzer 🌱 Step Outdoors (Touch Grass) 💻 Clone Local Python App
38
Crop Disease Classes
95.4%
Benchmark Accuracy
8.9 MB
MobileNetV2 Open Weights
100%
Local & Private (Zero Cloud)
Interactive Demonstration

Analyze Your Leaf Photo

📷

Upload or Drag & Drop Leaf Photo

Supports JPG, JPEG, and PNG (Good natural light, neutral backdrop)

Uploaded leaf preview
Or try built-in sample leaves:
🍅 Healthy Tomato 🥀 Tomato Early Blight 🍏 Apple Scab
Ready to Screen

Awaiting Leaf Photo

Model Statistical Confidence: -- %
Target Crop: --
Category: --
Latency: Local CPU

🩺 Cautious Care Recommendations:

  • 🪴 Select a sample or upload a photo to generate practical, organic next steps.
  • 💧 Always check soil moisture 2 inches down before applying treatments.
🌱 Step Outdoors
🌿 Core Challenge Focus

Touch Grass:
Connect with Living Nature

Screens are helpful for initial patterns, but genuine plant health happens out in the dirt. Step into natural sunlight, inspect living foliage, and nurture real soil.

Outdoor Care Checklist:

Inspect leaf underside for spores, mites, or ladybugs
Perform the 2-inch finger soil moisture test
Check canopy airflow and space between plants
Sanitize pruners with 70% alcohol before trimming
Breathe fresh air and enjoy 10 minutes outdoors
Privacy Guaranteed

Local Plant Health Journal

All scans and observation logs are stored directly in your local browser storage (and in local SQLite data/journal.db when running the Python desktop app). Zero personal data is sent to external servers.

Recent Observation Entries

Timestamp Plant / Crop Condition Statistical Confidence Outdoor Status
100% Open-Source AI

Model Specifications & Architecture

🧠 MobileNetV2 Backbone

Trained on the open PlantVillage dataset. Only ~3.5 million parameters and 8.9 MB weight footprint. Executes locally on CPU without dedicated GPU.

📜 Multi-Tier Open Licenses

Model Weights: Apache 2.0 (Permissive)
Training Dataset: CC-BY-SA 4.0
Project Source Code: MIT License

🛡️ Privacy & Independence

Commercial apps charge $30–$60/year and collect private telemetry. PlantGuard AI is completely free, open-source, and offline-capable.

Developer Setup

Run the Python App Locally

Clone the official repository and launch the local Gradio dashboard: 

git clone https://github.com/lalit-oli-mohan-479/PlantGuard-AI---Open-Source-Plant-Disease-Detector.git
cd PlantGuard-AI---Open-Source-Plant-Disease-Detector
pip install -r requirements.txt
python app.py

To execute the full test suite (22 unit & integration tests): 

pytest -v
🎉 Grass Touched!
You stepped outside and tended to your living plants! Observation recorded.