Image Classification

Purpose

This is my final project for the Break Through Tech Machine Learning Fundamentals Course. I chose to work with the CIFAR-10 dataset to deepen my understanding of Convolutional Neural Networks (CNNs) and gain hands-on experience applying machine learning to real-world image classification tasks. The project involved thorough data preprocessing, architecture experimentation, and hyperparameter optimization to enhance model performance. Through this, I aimed to build a solid foundation in deep learning workflows and improve my practical coding skills using TensorFlow.

Contents

• DefineAndSolveMLProblem.ipynb: Jupyter notebook that defines the full image classification pipeline, including data preprocessing, CNN architecture, training, and evaluation using TensorFlow.
• Image_Classification_Model.keras: Saved TensorFlow model trained on the CIFAR-10 dataset, serialized in Keras format.

Datasets

The project uses the CIFAR-10 dataset provided by Keras.

Model Architecture & Approach

• The pixel values from the CIFAR-10 dataset are used directly as features; no additional feature engineering is needed.
• Input data is normalized by scaling pixel values to a [0, 1] range, which improves convergence during training.
• A Convolutional Neural Network (CNN) is used for classification.
• The architecture consists of 4 convolutional layers, each followed by batch normalization and ReLU activation functions.
• A final dense layer with softmax activation is used to output class probabilities for this multi-class classification task.
• The model is compiled using:
  • Loss Function: Sparse Categorical Crossentropy (due to integer labels),
  • Optimizer: Stochastic Gradient Descent (SGD),
  • Metric: Accuracy.
• The model is trained and evaluated using the test set.
• To improve generalization and performance:
  • Grid search is used for hyperparameter optimization.
  • Dropout and pooling layers are added to mitigate overfitting and improve accuracy.

Results

Metric	Training Set	Testing Set
Loss	0.5263	0.5650
Accuracy	81.55%	80.46%

The model shows consistent performance between training and testing, indicating good generalization without significant overfitting.

Deployment

Download the .keras file and load it in a python file to use.

import tensorflow as tf
model = tf.keras.models.load_model("Image_Classification_Model.keras")