Implementation of Convolutional Neural Networks on MNIST dataset with Tensorflow - Detailed Steps
- To fully grasp the implementation of CNNs in tensorflow, we'll learn by classifying the simple MNIST dataset with 10 output classes. The basic idea of CNNs are the same even when we scale this up to the CIFAR-10 dataset.
- This is an expert level technical guide on tensorflow based on "Deep MNIST for Experts" by Google https://www.tensorflow.org/get_started/mnist/pros , Keras is not used so we can fully understand the nuts and bolts of tensorflow and be able to customize and tweak it completely.
- The theory of convolutional neural networks are not covered here as it is beyond the scope of this guide, please do some googling to understand kernels / filter sizes, strides, paddings and max pooling in CNNs before proceeding.
by Kelvin Kong
- Github: https://github.com/kelvinAI
- Linkedin: https://linkedin.com/in/kelvinkong