Learn about image classification with CNNs, the difference between the binary and multi-class image classification models, and how to use transfer learning for image classification in PyTorch.

Binary and multi-class image classification

The number of classes

Binary classification model

Multi-class classification model

Convolutional layers for images

RGB, grayscale, or alpha?

Adding a new convolutional layer

Creating a sequential block

Working with pre-trained models

Save and load a model

Loading a pre-trained model

Image classification with ResNet

Image Classification with CNNs

Detect objects in images by predicting bounding boxes around them and evaluate the performance of object recognition models.

Bounding boxes

Object recognition

Image tensors

Drawing a bounding box

Evaluating object recognition models

Calculate IoU

Bounding boxes prediction

Calculate NMS

Object detection using R-CNN

Pre-trained model backbone

Classifier block

Box regressor block

Region network proposals with Faster R-CNN

Anchor generator

Faster R-CNN model

Define losses for RPN and R-CNN

Object Recognition

Learn about the three types of image segmentation (semantic, instance, and panoptic), their applications, and the appropriate machine learning model architectures to perform each of them.

Introduction to image segmentation

Options

Semantic segmentation

Instance segmentation

Panoptic segmentation

Segmentation types

Creating binary masks

Segmenting image with a mask

Instance segmentation with Mask R-CNN

Segmenting with pre-trained Mask R-CNN

Analyzing model output

Displaying soft masks

Semantic segmentation with U-Net

Building a U-Net: layers definitions

Building a U-Net: forward method

Running semantic segmentation

Setup up semantic masks

Overlay instance masks

Image Segmentation

Generate completely new images with Generative Adversarial Networks (GANs). Learn to build and train a Deep Convolutional GAN, and how to evaluate the quality and variety of its outputs.

Introduction to GANs

GANs intuition

Generator

Discriminator

Deep Convolutional GAN

Convolutional Generator

Convolutional Discriminator

Training GANs

Generator loss

Discriminator loss

Training loop

Evaluating GANs

Generating images

Fréchet Inception Distance

Wrap-up

Image Generation with GANs

In this course, you'll use PyTorch to discover image classification, object recognition, segmentation, and image generation. You'll work with both binary and multi-class image classification models, utilize pre-trained models for deep learning tasks, and master object detection with bounding boxes. Additionally, you'll delve into image segmentation, including semantic, instance, and panoptic types, with real-world applications. Finally, you'll explore Generative Adversarial Networks (GANs) and learn to assess the quality and diversity of the generated images.

This course on deep learning for images using PyTorch will equip you with the practical skills and knowledge to excel in image classification, object detection, segmentation, and generation. 

<h2>Classify images with convolutional neural networks (CNNs)</h2>

You'll apply CNNs for binary and multi-class image classification and understand how to leverage pre-trained models in PyTorch. With bounding boxes, you'll also be able to detect objects within an image and evaluate the performance of object recognition models.

<h2>Segment images by applying masks</h2>

Explore image segmentation, including semantic, instance, and panoptic segmentation, by applying masks to images and learn about the different model architectures needed for each type of segmentation. 

<h2>Generate images with GANs</h2>

Finally, you'll learn how to generate your own images using Generative Adversarial Networks (GANs). You'll learn the skills to build and train Deep Convolutional GANs (DCGANs) and how to assess the quality and diversity of generated images.

By the end of this course, you'll have gained the skills and experience to work with various image tasks using PyTorch models.

Intermediate Deep Learning with PyTorch

Apply PyTorch to images and use deep learning models for object detection with bounding boxes and image segmentation generation.

Deep Learning for Images with PyTorch

Deep Learning in Python

Assessing deforestation or urban sprawl by segmenting different land cover types in satellite images.

Medical Imaging: Identifying and isolating specific regions of interest, such as tumors, blood vessels, or organs in MRIs or CT scans.

Retail: Identifying and counting specific products on shelves for inventory management.

Autonomous vehicles: Identifying pedestrians, vehicles, and other obstacles for safe navigation and tracking them over time.

Urban planning: Analyzing urban areas, where buildings are segmented individually, and roads, trees, etc., are segmented categorically.

Surveillance: Monitoring crowded public spaces by distinguishing between individual people while also identifying different areas or zones.

Segmentation types

Deep Learning for Images with PyTorch

Hands-on interactive exercise