damage_assessment_pipeline

AI-powered tool for spidermite damage assessment in tropical grasses

Overview

The damage_assessment_pipeline is the initial version of an AI-powered tool designed for assessing spidermite damage in tropical grasses. This software uses advanced image processing and machine learning models to perform damage classification for field images and plant segmentation using a salient object detection approach. Further refinement of models and fine-tuning is being actively researched. A deployed Gradio web application version is available on Hugging Face for online use, and users can also run the tool locally. An additional tool for batch processing is also included as a desktop app.

Features

• AI-powered assessment: Analyze images for spidermite damage (classification and plant segmentation).
• Web application: Use the tool directly through a Gradio web app on Hugging Face.
• Local execution: Run the tool locally for offline use.

Getting Started

1. Online Usage (Gradio Web App)

Access the deployed web app on Hugging Face: Damage Assessment Web App

2. Local Execution (Gradio Web App)

Prerequisites

• Python 3.12 or higher
• Git
• Anaconda / Miniconda

Dependencies

This project uses models and code from https://github.com/xuebinqin/U-2-Net

Steps

1. Clone the repository:

   git clone https://github.com/afruizh/damage_assessment_pipeline
   cd damage_assessment_pipeline/webapp

2. Install dependencies:

   pip install -r requirements.txt

3. Run the tool locally:

   python main.py

4. Open your browser and navigate to http://localhost:7860 to access the app.

2. Local Execution (Desktop App)

Prerequisites

• Python 3.12 or higher
• Git
• Anaconda / Miniconda

Dependencies

This project uses models and code from https://github.com/xuebinqin/U-2-Net

Steps

1. Clone the repository:

   git clone https://github.com/afruizh/damage_assessment_pipeline
   cd damage_assessment_pipeline/app

2. Install dependencies:

   pip install -r requirements.txt

3. Run the tool locally:

   python gui.py

4. The graphical user interface should appear

License

This project is licensed under the Apache-2.0 license.

Authors

Tropical Forages Team

Alliance Bioversity International & CIAT