This tutorial is based on my repository pytorch-computer-vision which contains PyTorch code for training and evaluating custom neural networks on custom data. By the end of this tutorial, you should be able to:

• Design custom 2D and 3D convolutional neural networks in PyTorch;
• Understand image dimensions, filter dimensions, and input dimensions;
• Understand how to choose kernel size, stride, and padding;
• Understand the basics of object-oriented programming relevant to PyTorch;
• Understand how to use nn.Sequential for better model organization.

Step 1: Download the code.

The repository with all the code is https://github.com/rachellea/pytorch-computer-vision

If you want to follow along with this tutorial and/or use the code, you…

At the end of this tutorial you should be able to:

• Load randomly initialized or pre-trained CNNs with PyTorch torchvision.models (ResNet, VGG, etc.)
• Select out only part of a pre-trained CNN, e.g. only the convolutional feature extractor
• Automatically calculate the number of parameters and memory requirements of a model with torchsummary

Predefined Convolutional Neural Network Models in PyTorch

There are many pre-defined CNN models provided in PyTorch, including:

• VGG family, named after the Visual Geometry Group at the University of Oxford. VGG models won first and second place in the localization and classification tasks, respectively, in the ImageNet ILSVRC-2014 competition. “VGG-N” has N layers. …

This post describes best practices for organizing machine learning projects that I have found to be highly effective during my PhD in machine learning.

Python

Python is a great language for machine learning. Python includes a bunch of libraries that are super useful for ML:

• numpy: n-dimensional arrays and numerical computing. Useful for data processing.
• pandas: data analysis library. pandas dataframes are essentially numpy arrays with descriptive strings as row and column labels. They can be sorted, filtered, grouped, concatenated, etc. Useful for data processing.
• PyTorch: build neural networks. Includes many pretrained models and clean computer vision datasets. PyTorch is my…

This post delves in to the use of CT scans in the COVID-19 pandemic, including current guidelines from medical experts (as of August 2020) and examples of recent research papers that use machine learning to make predictions from CT scans of COVID-19 patients.

Disclaimer: Nothing in this post is medical advice.

Diagnosis of COVID-19 with RT-PCR

The gold standard for diagnosis of COVID-19 is reverse transcription polymerase chain reaction (RT-PCR), which is a laboratory test that detects genetic material (RNA) from the COVID-19 virus:

This post provides an overview of chest CT scan machine learning organized by clinical goal, data representation, task, and model.

A chest CT scan is a grayscale 3-dimensional medical image that depicts the chest, including the heart and lungs. CT scans are used for the diagnosis and monitoring of many different conditions including cancer, fractures, and infections.

Clinical Goal

The clinical goal refers to the medical abnormality that is the focus of the study. The following figure illustrates some example abnormalities, shown as 2D axial slices through the CT volume:

Convolutional neural networks (CNNs) are the most popular machine leaning models for image and video analysis.

Example Tasks

Here are some example tasks that can be performed with a CNN:

• Binary Classification: given an input image from a medical scan, determine if the patient has a lung nodule (1) or not (0)
• Multilabel Classification: given an input image from a medical scan, determine if the patient has none, some, or all of the following: lung opacity, nodule, mass, atelectasis, cardiomegaly, pneumothorax

How CNNs Work

In a CNN, a convolutional filter slides across an image to produce a feature map (which is labeled “convolved feature” in…

This post offers the clearest explanation on the web for how the popular metrics AUC (AUROC) and average precision can be used to understand how a classifier performs on balanced data, with the next post focusing on imbalanced data. This post includes numerous simulations and AUROC/average precision plots for classifiers with different properties. All code to replicate the plots and simulations is provided on GitHub.

First, here is a brief intro to AUROC and average precision:

AUROC: Area Under the Receiver Operating Characteristic

The AUROC indicates whether your model can correctly rank examples. The AUROC is the probability that a randomly selected positive example has a higher…

Grad-CAM is a popular technique for visualizing where a convolutional neural network model is looking. Grad-CAM is class-specific, meaning it can produce a separate visualization for every class present in the image: