The Deep Learning Masterclass: Classify Images with Keras

Learn Python basics—variables, data types, operations, collections, control flow, functions, and classes—and how they power machine learning projects with Python.

Learn Python variable syntax and the four basic types: integers, floats, booleans, and strings—and use type() checks. Practice print statements with examples like age, rage factor, Hulk form, and name.

Explore Python variable operations, including assignment, arithmetic, modulus, and floor division, and master type conversion, printing results, and basic comparisons (>, <, ==, !=, is) with booleans, strings, and integers.

Explore Python collections by examining tuples and lists, learn indexing, length, and common operations like append, remove, and pop, and preview dictionaries as key-value stores.

Master Python control flow with while loops and for loops. Learn break and continue usage, avoid infinite loops, and iterate over lists efficiently.

Define and execute functions in Python, call functions, and encapsulate code; the lecture uses a Hulk rage example and explains global versus local variables.

Learn how to define and call Python functions with parameters and return values, pass input amounts, and capture true/false or multiple outputs from a function.

Learn about subclasses and superclasses, how inheritance shares attributes and behavior, and how to override with new initializers and methods, illustrated by Hulk and Avengers.

Explore Python basics across variables, types, collections (tuples, lists, dictionaries), control flow (if, while, for), functions, and classes with inheritance.

Explore the foundations of machine learning, neural networks, and convolutional neural networks, with three sections of video lectures and text notes on convolutions and how these networks work.

Define machine learning as a computer program that improves with data. Outline the four steps: processing data, building and training a model, and testing, highlighting supervised learning and image classification.

Explore how neural networks—networks of interconnected nodes with weights and biases across layers—recognize patterns and perform image classification, with training that adjusts weights to map inputs to outputs.

Explore how convolutional neural networks use convolution and max pooling to extract image features before dense layers. Understand image data as 3 color channels and how kernels shape feature maps.

Discover TensorFlow and Keras, compare their roles, and learn how to install them and use Keras to build, train, and test neural network models.

Explore Keras syntax via a sequential model to add convolutional, max pooling, dense, and dropout layers, then compile with an optimizer and loss and train with fit, evaluate, and predict.

Explore activation functions in neural networks, from identity and binary step to logistic, tanh, relu and leaky relu, and apply softmax for final image classification.

Explore how to load, format, and visualize CIFAR-10 data for image classification with Keras, and learn techniques to format inputs, labels, and model outputs.

Explore the cifar-10 dataset of 60,000 32x32 color images across 10 classes. Learn to access 50,000 training and 10,000 testing images with aligned labels for model training.

Understand how 32 by 32 rgb images and their train and test labels are stored as arrays, mapped to ten categories, visualized with matplotlib, and converted with keras to_categorical.

Learn how to prepare image data for Keras, including one-hot label encoding, normalization of train and test images, and flattening images, to enable effective image classification model building.

Build a convolutional neural network from scratch in Keras, layering convolutional, pooling, flatten, and dense layers with dropout and max-norm constraints to classify images.

Import and format inputs, then compile and train the model with a SGD optimizer and cross-entropy loss. Train with batch size 32 for 5–10 epochs and monitor loss and accuracy.

Explore gradient descent and optimizer concepts in Keras, including stochastic, batch, and mini-batch methods, and momentum, then see how backpropagation updates weights to reduce loss and improve accuracy.

Learn how to save and load pre-trained models using h5 files, evaluate and predict with the model, and freeze and save graphs for mobile imports.

Learn how to save and load a trained Keras model using an H5 file, including saving both architecture and weights in one step and loading for evaluation or prediction.