🍎 Fruit Freshness Detector (Prototype)

📌 Overview

Fruit Freshness Detector is a prototype computer vision project that estimates the freshness of fruit images using image variance analysis.

The system first identifies whether the uploaded image belongs to the fruit category using a Zero-Shot Image Classification model from Hugging Face. Once a fruit is detected, the image is analyzed using statistical image features to estimate its freshness level.

The project was developed as a learning-focused AI and Computer Vision prototype to explore image classification, feature extraction, and rule-based decision making.

---

🎯 Project Objective

The primary objectives of this project are:

• Detect whether an uploaded image contains a fruit.
• Extract visual features from the image.
• Estimate fruit freshness using image variance.
• Generate a simple freshness report.
• Explore the practical application of computer vision techniques.

---

🛠 Technologies Used

• Python
• Google Colab
• Hugging Face Transformers
• Zero-Shot Image Classification
• NumPy
• Pillow (PIL)

---

🔄 Project Workflow

Image Upload
      ↓
Fruit Classification
      ↓
Variance Calculation
      ↓
Freshness Estimation
      ↓
Freshness Report

---

📸 Sample Outputs

Fresh Apple Detection

Rotten Apple Detection

Fresh Banana Detection

---

🧠 Methodology

1. Fruit Classification

The uploaded image is passed through a Hugging Face Zero-Shot Image Classification model.

Candidate labels:

• Fruit
• Vegetable
• Baked
• Meat
• Dairy

Only images classified as Fruit proceed to freshness analysis.

---

2. Feature Extraction

The image is converted into a NumPy array and normalized.

Variance is calculated from the pixel values:

variance = np.var(img_array)

Variance measures the spread of pixel intensities within an image.

---

3. Freshness Estimation

Freshness is determined using predefined variance thresholds.

Variance Range	Freshness
Less than 0.04	Fresh
0.04 - 0.08	Okay
Greater than 0.08	Avoid

---

📊 Test Results

Fruit	Condition	Variance	Prediction
Apple	Fresh	0.033	Fresh
Apple	Rotten	0.102	Avoid
Banana	Fresh	0.029	Fresh
Banana	Rotten	0.081	Avoid
Orange	Fresh	0.069	Okay
Orange	Rotten	0.102	Avoid
Grapes	Fresh	0.070	Okay
Pear	Fresh	0.053	Okay
Pear	Rotten	0.072	Okay

---

📈 Key Observations

During testing, the system performed well for:

• Apples
• Bananas

The system showed mixed results for:

• Pears
• Grapes
• Oranges

This demonstrates that image variance can be useful for freshness estimation but is not sufficient as a standalone feature for all fruit types.

---

⚠️ Limitations

• Prototype-level implementation.
• Supports fruit freshness estimation only.
• Relies solely on image variance.
• Performance depends on image quality and lighting conditions.
• Different fruit textures affect variance values.
• Not intended for commercial deployment.

---

🚀 Future Improvements

Potential future enhancements include:

• Training a CNN-based Fresh vs Rotten classifier.
• Using a dedicated fruit freshness dataset.
• Adding support for vegetables.
• Combining multiple image features instead of variance alone.
• Building a Streamlit web application.
• Deploying the project on the cloud.

---

📚 Learning Outcomes

This project helped in understanding:

• Computer Vision Fundamentals
• Image Classification
• Feature Extraction
• NumPy Image Processing
• Hugging Face Transformers
• Rule-Based Classification Systems
• Model Evaluation and Testing

---

💡 Conclusion

This project successfully demonstrates a prototype approach for fruit freshness estimation using image classification and variance-based analysis.

Although the model has limitations, it serves as a strong proof-of-concept and provides valuable insights into computer vision workflows, image processing, and AI-based decision making.

---

👨‍💻 Author

Prayansh Gupta

B.Tech – Artificial Intelligence & Data Science

Galgotias College of Engineering & Technology (2023-27)

Prototype project developed for learning and experimentation in Computer Vision and Artificial Intelligence.