Real-World Python Neural Nets Tutorial - Image Classification with CNN - Tensorflow & Keras

Learn to build and train a convolutional neural network (CNN) for image classification using TensorFlow and Keras in this comprehensive tutorial. Walk through the entire process of creating a CNN to classify images of rock, paper, and scissors. Begin with dataset acquisition and preparation, including converting data to NumPy format and normalizing values. Progress to training an initial network, then explore CNN approaches and GPU acceleration on Google Colab. Enhance the model by implementing techniques such as image size reduction, max pooling, and dropout. Utilize Kerastuner for automatic hyperparameter optimization. Cover model saving and loading, plotting NumPy arrays as images, and converting JPG/PNG images to NumPy format. Gain practical experience in applying deep learning techniques to real-world image classification problems.

Syllabus

Video Overview
Getting Started Setup & Installation
Finding datasets to use
Data Preparation
Additional Data Prep Convert data to NumPy format
Reshape Data & Normalize values between 0-1
Train our first network to classify images
Convolutional Neural Net CNN approach
Using GPU on Google Colab speed up training
Improving our CNN reduce image size, max pooling, dropout, etc
Using Kerastuner to automatically pick best hyperparameters
Save & Load our models
Plot NumPy arrays as images
Convert JPG/PNG images to NumPy
Final thoughts