CIFAR-10 Image Classification - CNN with TensorFlow

Personal Projects #Python#Machine Learning#Data Science Featured

Overview#

A convolutional neural network (CNN) built with TensorFlow and Keras to classify images from the CIFAR-10 dataset into 10 categories: Airplane, Automobile, Bird, Cat, Deer, Dog, Frog, Horse, Ship, and Truck.

Key Achievements#

Achieved 84% training accuracy and 80% vaidation accuracy
Model correctly classified all test images across different categories
Demonstrated effective regularization with Dropout preventing overfitting
Achieved average loss of 0.62 after 30 training epochs

Implementation#

Data Preprocessing: Normalized pixels to [0,1], encoded labels
CNN Architecture: Input (32x32x3) -> Progressive Conv2D layers (32->64->128) with MaxPool and Dropout -> Flatten -> Dense(512) -> Dense(10) with Softmax
Regularization: Dropout (0.25 in conv layers, 0.5 before output) to prevent overfitting
Training: Adam optimizer with categorical cross-entropy loss over 30 epochs

Accuracy & Loss