170 lines
4.6 KiB
Markdown
170 lines
4.6 KiB
Markdown
|
|
🌱 Plant Health & Identification API
|
|
|
|
An intelligent backend service built for our hackathon project. This API uses a dual-model AI system to analyze an image of a plant, first identifying its species and then assessing its health.
|
|
|
|
🚀 The Problem
|
|
|
|
Have you ever wondered what kind of plant you have or why its leaves are suddenly turning yellow? Our project aims to provide a simple, accessible answer. This repository contains the backend API that powers our application, capable of providing a comprehensive plant analysis from a single image.
|
|
|
|
🧠 The AI Pipeline
|
|
|
|
This API uses a two-stage process to analyze an image:
|
|
|
|
Species Identification: We use a custom-trained TensorFlow/Keras model, built on a MobileNetV2 architecture. This model was fine-tuned on a dataset of six specific plant types to achieve high accuracy in identifying which plant is in the image.
|
|
|
|
Health Assessment: Once the plant's species is known (e.g., "tomato"), we use OpenAI's powerful, open-source CLIP model. We dynamically generate text prompts like "a photo of a healthy tomato plant" or "a photo of a sick tomato plant with yellow spots" and ask CLIP which description best matches the image. This zero-shot approach allows for a flexible and nuanced understanding of the plant's health.
|
|
|
|
✨ Features
|
|
|
|
Identify 6 Plant Species: Accurately distinguishes between tomato, basil, mint, lettuce, rosemary, and strawberry.
|
|
|
|
Assess 4 Health States: Classifies plants as Healthy, Diseased, Dehydrated, or Dead.
|
|
|
|
Confidence Scores: Provides confidence levels for both the species identification and the health assessment.
|
|
|
|
Simple JSON API: Easy to integrate with any frontend or mobile application.
|
|
|
|
🛠️ Tech Stack
|
|
|
|
Backend Framework: Flask
|
|
|
|
AI / Machine Learning: TensorFlow (Keras), PyTorch, OpenAI CLIP
|
|
|
|
Image Processing: Pillow, OpenCV
|
|
|
|
Core Language: Python 3.10
|
|
|
|
🔌 API Documentation
|
|
|
|
This is the documentation for the main analysis endpoint.
|
|
|
|
Analyze a Plant Image
|
|
|
|
Endpoint: /analyze
|
|
|
|
Method: POST
|
|
|
|
Body: multipart/form-data
|
|
|
|
The request must contain a file field named image.
|
|
|
|
Successful Response (Status 200 OK)
|
|
|
|
The API will return a JSON object with the analysis.
|
|
|
|
Example Response:
|
|
|
|
Generated json
|
|
{
|
|
"plant_species": "tomato",
|
|
"identification_confidence": "97.45%",
|
|
"health_status": "Healthy",
|
|
"health_confidence": "89.12%",
|
|
"health_breakdown": {
|
|
"Healthy": 0.8912,
|
|
"Diseased": 0.0562,
|
|
"Dehydrated": 0.0421,
|
|
"Dead": 0.0105
|
|
}
|
|
}
|
|
|
|
|
|
Field Descriptions:
|
|
|
|
Key Type Description
|
|
plant_species String The identified species of the plant.
|
|
identification_confidence String The model's confidence in the species identification.
|
|
health_status String The most likely health status of the plant.
|
|
health_confidence String The model's confidence in the health assessment.
|
|
health_breakdown Object A dictionary of raw probability scores for each health state.
|
|
Error Responses
|
|
|
|
If the request is missing an image, the API will return:
|
|
|
|
Status: 400 Bad Request
|
|
|
|
Body: {"error": "No image file provided"}
|
|
|
|
For any other server-side issues, the API will return:
|
|
|
|
Status: 500 Internal Server Error
|
|
|
|
Body: {"error": "An internal server error occurred."}
|
|
|
|
🖥️ How to Run Locally
|
|
|
|
To run this backend server on your own machine, follow these steps.
|
|
|
|
Clone the Repository:
|
|
|
|
Generated bash
|
|
git clone <repository-url>
|
|
cd <repository-name>/backend
|
|
IGNORE_WHEN_COPYING_START
|
|
content_copy
|
|
download
|
|
Use code with caution.
|
|
Bash
|
|
IGNORE_WHEN_COPYING_END
|
|
|
|
Create a Virtual Environment:
|
|
|
|
Generated bash
|
|
python3 -m venv venv
|
|
source venv/bin/activate
|
|
IGNORE_WHEN_COPYING_START
|
|
content_copy
|
|
download
|
|
Use code with caution.
|
|
Bash
|
|
IGNORE_WHEN_COPYING_END
|
|
|
|
Install Dependencies:
|
|
This can take a while as it will download TensorFlow and PyTorch.
|
|
|
|
Generated bash
|
|
pip install -r requirements.txt
|
|
IGNORE_WHEN_COPYING_START
|
|
content_copy
|
|
download
|
|
Use code with caution.
|
|
Bash
|
|
IGNORE_WHEN_COPYING_END
|
|
|
|
Place the Model:
|
|
Make sure you have the trained Keras model (BestModel.keras) inside the models/ directory.
|
|
|
|
Run the Server:
|
|
|
|
Generated bash
|
|
python app.py
|
|
IGNORE_WHEN_COPYING_START
|
|
content_copy
|
|
download
|
|
Use code with caution.
|
|
Bash
|
|
IGNORE_WHEN_COPYING_END
|
|
|
|
The API will now be running on your local machine at http://127.0.0.1:5000.
|
|
|
|
📁 Project Structure
|
|
Generated code
|
|
/backend
|
|
|-- app.py # The main Flask server and API logic.
|
|
|-- models/ # Folder for the trained Keras model.
|
|
| |-- BestModel.keras
|
|
|-- requirements.txt # Python dependencies.
|
|
|-- .gitignore # Files to be ignored by Git (like the venv).
|
|
IGNORE_WHEN_COPYING_START
|
|
content_copy
|
|
download
|
|
Use code with caution.
|
|
IGNORE_WHEN_COPYING_END
|
|
👥 Authors
|
|
|
|
[Your Name]
|
|
|
|
[Teammate's Name]
|
|
|
|
[Teammate's Name]
|