Introduction to Deep Learning for Computer Vision

Overview

Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!

Grab it

Starting with zero deep learning knowledge, this foundational course will guide you to effectively train cutting-edge models for image classification purposes. From analyzing medical images to recognizing traffic signs, classification is important for many applications. Classification models also serve as the backbone for more complicated object detection models. Through hands-on projects, you will train and evaluate models to classify street signs and identify the letters of American Sign Language. By completing this course, you will develop a strong foundation in deep learning for image analysis and will be equipped with the skills to tackle real-world computer vision challenges. By the end of this course, you will be able to: • Explain how deep learning networks find image features and make predictions • Retrain common models like GoogLeNet and ResNet for specific applications • Investigate model behavior to identify errors and determine potential fixes • Improve model performance by tuning hyperparameters • Complete the entire deep learning workflow in a final project For the duration of the course, you will have free access to MATLAB, software used by top employers worldwide. The courses draw on the applications using MATLAB, so you spend less time coding and more time applying deep learning concepts.

Syllabus

Introduction to Deep Learning with Images

Learn the key components of convolutional neural networks and train a simple classification model

Transfer Learning

Retraining networks with new data is the most common way to apply deep learning in industry. In this module, you'll retrain common networks, set appropriate values for training options, and compare results from different models.

Investigating Network Behavior

Explaining how models make predictions is increasingly important. In this module, you'll use confidence scores and visualizations to determine what regions of an image the model is using to make predictions. You'll also identify common errors and adjust training options to improve performance.