HafizRehan5833/fruit-freshness-detection

🍏 FruitGuard AI — Fruit Freshness Detection Model

This repository hosts the official trained Deep Learning model for FruitGuard AI, a high-precision computer vision model classifying fruits as Fresh or Spoiled with 99.84% accuracy.

• GitHub Repository: HafizRehan5833/system7_project
• Model Architecture: MobileNetV2 (Transfer Learning from ImageNet)
• Input Resolution: 224 × 224 RGB
• Classes: fresh (0), spoiled (1)
• Supported Varieties: Apple, Banana, Grape, Guava, Jujube, Orange, Pomegranate, Strawberry

---

📊 Evaluation & Benchmark Results

Evaluated on 1,850 completely held-out test images:

Metric	Fresh Fruit	Spoiled Fruit	Overall System
Accuracy	—	—	99.84% (1,847 / 1,850)
Precision	99.68%	100.00%	99.84%
Recall	100.00%	99.67%	99.84%
F1-Score	0.9984	0.9984	0.9984

---

💻 Quickstart: How to Use

1. Download & Load the Model

import numpy as np
from PIL import Image
import tensorflow as tf
from huggingface_hub import hf_hub_download

# Download weights from Hugging Face
model_path = hf_hub_download(
    repo_id="HafizRehan5833/fruit-freshness-detection",
    filename="fruit_freshness_model.keras"
)

# Load the Keras model
model = tf.keras.models.load_model(model_path, compile=False)

def predict_freshness(image_path: str):
    # Preprocess: RGB, 224x224
    image = Image.open(image_path).convert("RGB").resize((224, 224))
    batch = np.expand_dims(np.asarray(image, dtype=np.float32), axis=0)

    # In-graph rescaling is handled automatically by the model
    spoiled_probability = float(model.predict(batch, verbose=0).ravel()[0])
    fresh_probability = 1.0 - spoiled_probability

    label = "spoiled" if spoiled_probability >= 0.5 else "fresh"
    confidence = spoiled_probability if label == "spoiled" else fresh_probability

    return {
        "verdict": label,
        "confidence": round(confidence * 100, 2),
        "fresh_probability": round(fresh_probability, 4),
        "spoiled_probability": round(spoiled_probability, 4),
    }

# Example run:
# print(predict_freshness("test_apple.jpg"))

---

🏭 Real-World Applications

• Automated Conveyor Sorting: Optical inspection in packhouses with pneumatic rejectors.
• Retail Quality Control: Automated supermarket shelf monitoring and dynamic markdown pricing.
• Cold Chain Logistics: Receiving dock QA verification to prevent cross-contamination.
• Food Processing: Pre-crush filter for juices, puree, and baby food manufacturing.

For the complete technical specifications and enterprise architecture, visit the GitHub Repository.