🧠 Object Recognition on CIFAR-10

This project explores various machine learning and deep learning techniques for object recognition using the CIFAR-10 dataset — a benchmark dataset of 60,000 32x32 color images across 10 categories.

🔗 Project Website
📂 CIFAR-10 Dataset

---

📌 Overview

The goal of this project is to evaluate and compare different classification algorithms on the CIFAR-10 dataset. We explore both traditional ML models and deep learning architectures, emphasizing the role of preprocessing and feature extraction in handling high-dimensional image data.

---

🛠️ Methods Implemented

🔍 Preprocessing & Feature Engineering

• Data Normalization
• Dimensionality Reduction using PCA

🧪 Machine Learning Models

• Linear Regression
• K-Nearest Neighbors (KNN)
• Logistic Regression
• Support Vector Machines (SVM - Linear & RBF)
• Decision Trees
• Naive Bayes
• Gaussian Mixture Models (GMM)
• DBSCAN
• K-Means Clustering

🤖 Deep Learning Models

• Artificial Neural Networks (ANN)
• Convolutional Neural Networks (CNN)

---

📊 Results

Model	Accuracy (%)
Linear Regression	10.96
K-Nearest Neighbors (KNN)	39.98
Logistic Regression	40.28
SVM (Linear Kernel)	38.47
SVM (RBF Kernel)	42.10
DBSCAN	30.64
Naive Bayes (Gaussian)	36.98
Gaussian Mixture Model (GMM)	36.50
Decision Tree	27.28
K-Means Clustering	22.69
Artificial Neural Network (8 layers, L2 reg.)	52.00
Convolutional Neural Network (Basic CNN)	80.50

📌 The CNN model significantly outperforms traditional methods, underscoring:

• The power of deep learning in image recognition tasks.
• The importance of feature extraction and representation in high-dimensional spaces.

---

💡 Key Takeaways

• Dimensionality reduction (PCA) improves traditional model performance but may still fall short on complex image data.
• Deep learning models, especially CNNs, are highly effective for image classification due to their hierarchical feature learning.

---

🚀 Getting Started

To run the code:

1. Clone the repository.

2. Install dependencies using:

   pip install -r requirements.txt

3. Run the provided notebooks/scripts to train and evaluate models.

---