KwokHing / TF2-Cifar10-CNN-Demo

Demo on performing multiclass image classification using Convolutional Neural Network (CNN) in Tensorflow 2. Techniques such as earlystopping, batchnormalizing and dropout are explored to prevent overfitting

Tensorflow 2 CNN Cifar10 Classification

This demo deploys the use of Convolutional Neural Networks (CNN) in Tensorflow 2 to classify Cifar10 images.

• Tensorflow Data Pipeline
• Convolutional Neural Networks (CNN)
• Techniques that helps prevents overfitting (EarlyStopping, BatchNormalization, Dropout)
• Tensorboard
• Saving Model (.h5, tf, weights)

Getting started

Open TF2_Cifar10_CNN.ipynb on a jupyter notebook environment, or Google colab. The notebook consists of further technical details.

Future Improvements

• Explore the use of data augmentation in image classifcation
• Explore the model performance on Cifar100 dataset