Fine tuning the VGG16 deep convolutional neural network to classify images from the CIFAR-10 dataset.
Refer to vgg16_for_CIFAR10.ipynb
for step by step explanations and discussion.
This was a personal project, to familiarise myself with Keras, Tensorflow, Google Colab and image processing in general. I initially began working with Jupyter Notebooks and created a fairly shallow network for simple data classification - however, it was too simple to classify images. Therefore, using help from video tutorials online I learnt how to fine tune and manipulate pre-existing deep networks, and set myself the challenge of classifying the CIFAR-10 dataset with a modified VGG16 model.
The modified networked managed an accuracy of just over 70%, and with better hardware and memory storage, the model should perform even better.