🧠 Image Classification with CNN and Transfer Learning

This project explores image classification on the CIFAR-10 dataset using two approaches:

1. A Custom CNN built from scratch
2. Transfer Learning with pre-trained models

We compare performance, training times, and trade-offs between the two approaches.

---

📂 Project Structure

• ProjectCifar.ipynb → Main notebook with all experiments
• README.md → Project documentation (this file)

---

📊 Dataset

• CIFAR-10:
  • 60,000 color images (32×32 pixels)
  • 10 categories: airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck
  • Train set: 50,000 images, Test set: 10,000 images

---

⚙️ Methods

1. Custom CNN

   • Several convolutional + pooling layers
   • Flatten → Dense → Output
   • Trained from scratch

2. Transfer Learning

   • Used pre-trained models (e.g., MobileNet with different training layers)
   • Fine-tuned on CIFAR-10 dataset
   • Faster convergence and improved accuracy

📦 Requirements

Install dependencies with:

pip install -r requirements.txt