South American Flag Classification with VGG16 + XGBoost

Project Overview

This project performs highly accurate image classification of South American country flags using a hybrid deep learning approach. I utilize VGG16 for feature extraction and XGBoost for classification, achieving 97% test accuracy on a custom dataset of over 10,000 flag images.

---

Key Features

• Hybrid architecture combining VGG16 with XGBoost
• 97% test accuracy on South American flags
• Comprehensive training and evaluation pipeline
• Detailed performance visualizations

---

Model Architecture

graph LR
    A[Input Flag Image] --> B[VGG16 Feature Extraction]
    B --> C[Remove Final Classification Layer]
    C --> D[Extracted Feature Vector]
    D --> E[XGBoost Classifier]
    E --> F[Predicted Country]

Loading

---

Accuracy Metrics

Metric	Value
Accuracy	97.0%
Dataset Size	~10,000 images
Classes	12 South American countries

---

Confusion Matrix

Training Progress

Accuracy Progress	Loss Reduction

---

---

Notes:

- Model optimized specifically for South American flags  
- Automatic image resizing during inference  
- Internet connection required for initial download of the model. 
- Trained model hosted on Hugging Face:  
  🔗 [flag_classifier_vgg16.h5 – 116 MB](https://huggingface.co/7mgppp/flag-classifier-vgg16-model/blob/main/flag_classifier_vgg16.h5)

---