Overview

Updated in May 2025. This course now features Coursera Coach — your interactive learning companion that helps you test your knowledge, challenge assumptions, and deepen your understanding as you progress. Master the power of neural networks with this hands-on deep learning course built entirely in PyTorch. Designed for data scientists, AI practitioners, and developers, this course guides you step by step through building, training, and evaluating models for image, audio, and sequence-based tasks using one of the industry’s most popular frameworks. You’ll begin by exploring classification models, learning how to handle binary and multi-class problems, interpret confusion matrices, and analyze ROC curves. Through practical exercises, you’ll prepare data, design dataset classes, and build your own neural network architectures to solve real classification challenges. Next, you’ll move into Convolutional Neural Networks (CNNs), where you’ll develop both image and audio classification systems. You’ll learn how CNN layers work, implement preprocessing pipelines, and construct models for binary and multi-class image tasks. You’ll also extend these skills to audio classification, giving you a broader understanding of how CNNs apply across domains. From there, you’ll dive into object detection, mastering accuracy metrics, labeling formats, and the YOLO (You Only Look Once) algorithm. Hands-on coding sessions walk you through data preparation, training, and inference so you can build complete, end-to-end detection workflows. In the final modules, you’ll explore neural style transfer, transfer learning with pre-trained networks, and sequence modeling using RNNs and LSTMs — gaining the skills to tackle advanced deep learning applications. By the end of this course, you will have: - Built and evaluated neural network models for binary and multi-class classification. - Designed and trained CNNs for image and audio data. - Implemented object detection workflows using YOLO. - Applied neural style transfer and leveraged pre-trained models for transfer learning. - Developed RNN and LSTM models for sequence-based tasks. - Gained the confidence to use PyTorch for real-world deep learning projects. This course is ideal for learners with experience in Python and a foundational understanding of machine learning and deep learning concepts who want to advance their skills in building neural networks with PyTorch.

Syllabus

Classification Models

In this module, we will delve into the realm of classification models, focusing on their types, evaluation metrics, and implementation. You will learn about key concepts such as the confusion matrix and ROC curve, and engage in practical exercises to build and evaluate multi-class classification models.

CNN: Image Classification

In this module, we will explore the power of convolutional neural networks (CNNs) in image classification tasks. You will learn about the CNN architecture, preprocess images for optimal results, and gain hands-on experience in implementing binary and multi-class image classification models.

CNN: Audio Classification

In this module, we will focus on using convolutional neural networks for audio classification. You will get a comprehensive introduction to the topic, learn how to conduct exploratory data analysis on audio data, and engage in practical exercises to build and evaluate your own audio classification models.

CNN: Object Detection

In this module, we will dive into object detection using convolutional neural networks. You will learn about essential accuracy metrics, implement popular object detection algorithms like YOLO, and utilize GPU resources for training and inference to build robust object detection models.

Style Transfer

In this module, we will cover the fascinating topic of neural style transfer. You will understand the underlying principles, implement style transfer algorithms through coding, and explore various creative applications to transform images in unique ways.

Pre-Trained Networks and Transfer Learning

In this module, we will delve into pre-trained networks and transfer learning. You will learn how to leverage pre-trained models, implement transfer learning techniques through coding exercises, and understand the advantages of applying these concepts to various machine learning tasks.

Recurrent Neural Networks

In this module, we will introduce recurrent neural networks (RNNs) and their applications. You will explore the basics of RNNs, implement Long Short-Term Memory (LSTM) networks through practical coding exercises, and engage in tasks designed to deepen your understanding of these powerful models.