Project Description

This application allows users to upload images or capture images from a camera, then uses a trained YOLO model to detect and classify corn kernels into categories (black, corn, white, and fragments). It provides count statistics for each category and maintains a cumulative count.

Documentation

Application Interface

The application provides a simple web interface where users can upload images or capture photos using their camera. The processed images are displayed along with classification results.

Detection Results

After processing, the application displays the analyzed image with detected corn kernels and a detailed breakdown of kernel counts by category:

Category	Description
Good Seeds	Healthy corn kernels with no infection
Aspergillus niger	Kernels infected with the Aspergillus niger fungus, a common mycotoxin-producing mold that appears as black spots
Penicillium	Kernels infected with Penicillium fungi, which typically appears as blue-green or black mold on corn
Corn Fragments	Broken or partial corn kernel pieces
Total Seeds	The total count of all detected kernels and fragments

The application also maintains a cumulative count of detected kernels across multiple images, which can be reset as needed.

Model Information

This application uses a custom-trained YOLOv8 model for corn kernel detection and classification. The model achieves impressive accuracy:

• mAP: 99.2%
• Precision: 98.2%
• Recall: 97.7%

Access the Model

You can access, duplicate, or experiment with our model on Roboflow:
Corn Detection Model

The dataset is available for download through the Roboflow platform, allowing you to train your own models or modify our approach for similar detection tasks.

Setup Instructions

Prerequisites

• Python 3.x
• Pip package manager

Installation

1. Clone the repository or download the project files

2. Create and activate a virtual environment (optional but recommended):

   python -m venv env
   source env/bin/activate  # On Windows: env\Scripts\activate

3. Install the required dependencies:

   pip install flask opencv-python ultralytics werkzeug

4. Ensure the YOLO model file (best.pt) is in the model directory

5. Make sure all required directories exist:

   mkdir -p check results static templates uploads

Configuration

You may need to update the file paths in app.py to match your system:

UPLOAD_FOLDER = '/path/to/your/project/check'
RESULTS_FOLDER = '/path/to/your/project/results'
STATIC_FOLDER = '/path/to/your/project/static'

Running the Application

1. Start the Flask server:

   python app.py

2. Open a web browser and navigate to:

   http://127.0.0.1:5000/
   

Usage

1. The application provides two ways to submit images:

   • Upload an image file
   • Capture an image using your camera

2. After processing, the application will:

   • Display the processed image with detections
   • Show counts for each kernel type (black, corn, white, fragments)
   • Display the total kernel count
   • Show the cumulative count across all processed images

3. You can reset the cumulative count using the reset button

Project Structure

• app.py: Main Flask application
• best.pt: YOLO model trained for corn kernel detection
• index.html: HTML template for the web interface
• check: Temporary storage for uploaded images
• results: Directory for saving processed images
• static: Static files served by the application
• uploads: Directory for storing uploaded files

Technologies Used

• Flask: Web framework
• OpenCV: Image processing
• Ultralytics YOLO: Object detection model
• JavaScript/HTML/CSS: Frontend

License

[agungferdi] thanks