🧠 Image Classification using CNN (CIFAR-10)

📌 Project Purpose

The purpose of this project is to build a Convolutional Neural Network (CNN) using TensorFlow and Keras to classify images from the CIFAR-10 dataset, which contains 60,000 32x32 color images in 10 different classes.

---

❓ Problem Statement

Traditional image classification approaches often struggle with low accuracy and high computational costs when working with large image datasets. This project aims to develop a deep learning model that can effectively learn spatial hierarchies and features in images to accurately predict the correct object class in unseen images.

---

🎯 Objectives

• Load and preprocess CIFAR-10 image data.
• Build a CNN model using TensorFlow/Keras.
• Train the model and evaluate its performance using validation and test sets.
• Use techniques like Dropout and Pooling to prevent overfitting and enhance accuracy.
• Visualize training/validation performance and predictions.

---

📊 Dataset: CIFAR-10

• 60,000 color images (32x32 pixels)

• 10 classes:

  • airplane
  • automobile
  • bird
  • cat
  • deer
  • dog
  • frog
  • horse
  • ship
  • truck

• Training set: 50,000 images

• Test set: 10,000 images

Dataset is available via keras.datasets.cifar10.

---

🛠️ Technologies Used

• Python
• TensorFlow / Keras
• NumPy, Matplotlib, Seaborn
• CNN (Conv2D, MaxPooling, Flatten, Dense)
• Google Colab / Jupyter Notebook

---

🧱 Model Architecture

• ✅ Input: 32x32x3 RGB images
• ✅ Conv2D + MaxPooling2D layers (multiple blocks)
• ✅ Flatten layer
• ✅ Dense layers with ReLU activation
• ✅ Dropout for regularization
• ✅ Output: Dense layer with 10 units and Softmax activation

---

📊 Model Summary

Model: "sequential"

Layer (type)	Output Shape	Param #
Conv2D	(32, 32, 32)	896
Conv2D	(32, 32, 32)	9,248
MaxPooling2D	(16, 16, 32)	0
Dropout	(16, 16, 32)	0
Conv2D	(16, 16, 64)	18,496
Conv2D	(16, 16, 64)	36,928
MaxPooling2D	(8, 8, 64)	0
Dropout	(8, 8, 64)	0
Flatten	(4096)	0
Dense	(512)	2,097,664
Dropout	(512)	0
Dense	(10)	5,130
Total Parameters		2,168,362
Trainable Parameters		2,168,362
Non-trainable Parameters		0

---

📈 Results

• ✅ Final Validation Accuracy: ~80%
• ✅ Final Test Accuracy: ~79%
• ✅ Model performs well across most CIFAR-10 classes

---

📸 Prediction Sample

• python:
• predict_image(index=5) # shows prediction and actual label with image

---

🖼️ Output:

• Predicted: cat | Actual: cat

---

🛠️ How to Run

1. Clone the repository:

git clone https://github.com/shivharebhupendra/Image-Classification-with-CNN.git cd Image-Classification-with-CNN

2. Install dependencies:

pip install -r requirements.txt

3. Run the training notebook:

• Open the Jupyter notebook or .py file.
• Execute cells to load data, build model, train and evaluate.

4. Predict on test images:

predict_image(12) # function to visualize prediction

---

⚙️ Requirements

• Python
• TensorFlow
• NumPy
• Matplotlib
• Jupyter Notebook

Install them via:

• pip install tensorflow numpy matplotlib

👨‍💻 Author

Bhupendra Shivhare

Aspiring Data Scientist | Deep Learning Enthusiast

LinkedIn: www.linkedin.com/in/bhupendra-shivhare-a8a02a25b

📧 Email: shivharebhupendra@gmail.com