GRADION

💡 Inspiration

GRADION started with a simple observation: onion quality grading often depends on manual visual inspection, which can be time-consuming and subjective.

We wanted to explore whether computer vision could make onion grading faster, more consistent, and scalable. This led us to develop GRADION, an AI-based system that evaluates onions based on their visible characteristics.

🛠️ How We Are Building It

GRADION uses computer vision to analyze images of onions and evaluate visual characteristics such as:

Size
Shape
Color
Visible defects
Overall appearance

The system processes an onion image, extracts relevant visual features, and uses an AI model to classify the onion into an appropriate quality grade.

Alongside the AI model, we are developing a simple application interface through which users can capture or upload onion images and receive the grading result.

Our development process involves:

Image Collection
       ↓
Data Cleaning & Labelling
       ↓
Image Preprocessing
       ↓
Model Training
       ↓
Model Evaluation
       ↓
Application Integration
       ↓
Onion Quality Grade

📚 What We Learned

Building GRADION taught us that developing an AI solution is much more than simply training a model.

We learned about:

Computer vision and image classification
Dataset collection and preparation
Image preprocessing and augmentation
Model training and evaluation
Building an application around an AI model
Designing a simple user workflow
Team collaboration and problem-solving
Converting a real-world problem into a technical solution

Most importantly, we learned that the quality of the data can be just as important as the model itself.

🚧 Challenges We Are Facing

One of our biggest challenges is obtaining a large, diverse, and reliable dataset of onion images.

Onions can look different depending on:

Variety
Lighting conditions
Camera angle
Background
Size and shape
Type and visibility of defects

A model trained only on controlled images may not perform reliably in real-world conditions. We are therefore working toward making our dataset more diverse and improving the robustness of the model.

Another challenge is deciding which visual characteristics should have the greatest influence on the final grade. We are experimenting with different approaches while keeping the system practical and easy to use.

🚀 Our Vision

Our goal with GRADION is to develop a fast, consistent, and scalable computer-vision-based onion grading system that can reduce dependence on subjective manual inspection.

As the project develops, we aim to improve the model with larger and more diverse datasets, better grading accuracy, and real-world testing.

GRADION is still a work in progress, but the challenges we have encountered are helping us understand what it takes to turn an AI concept into a practical solution.