Human Pose Estimation with Deep Neural Networks

Keypoint estimation, Object detection, R-CNN, Transfer learning, ResNet

Armin Kappeler and Aadit Patel

8 weeks · 8-10 hours per week

In this liveProject, you’ll take on the role of a machine learning engineer working for a company developing augmented reality apps. These apps include games, virtual shopping assistants, and fitness coaches that need to be able to reliably recognize the shape of a human body. Your challenge is to create an application for human pose estimation: detecting a human body in an image and estimating its key points such as knees and elbows. To do this, you’ll build a convolutional neural network from scratch, training your model using Google Colab and your GPU. At the end of this liveProject, you’ll have completed an interactive demo application that uses a simple webcam to detect and predict human keypoints.

project authors

Armin Kappeler

Armin Kappeler received his M.S. and Ph.D. degree in electrical engineering from Northwestern University in 2016 where he worked on different deep neural networks applications for Image and video classification and recovery. For his thesis, he applied deep neural networks to video super-resolution. In 2015, he joined Verizon Media Group (Yahoo Research) as a research engineer, where he is working on image captioning, active learning, human pose estimation and other related computer vision applications.

Aadit Patel

Aadit Patel received a B.S. in aerospace engineering and an M.S. in computer science, both from the University of California, Los Angeles (UCLA), where his primary focuses included autonomous systems and reinforcement learning. He has several years of industry experience as a research engineer and data scientist, including at Boeing, Yahoo, and Flyr Labs. Currently, Aadit is a lead data scientist at The Not Company, where he works on custom deep learning architectures to generate novel plant-based foods.

Prerequisites

This liveProject is for intermediate Python programmers who are familiar with machine learning. Knowledge of PyTorch and NumPy will be helpful. To begin this liveProject, you will need to be familiar with:

TOOLS

Basics of PIL
Basics of JSON
Basics of Matplotlib
Intermediate PyTorch
Intermediate NumPy

TECHNIQUES

Intermediate machine learning concepts such as classification and regression
Basics of matrix and vector operations

you will learn

In this liveProject, you’ll learn how to build deep neural networks, and how to utilize them for computer vision. The building blocks of this project are easily transferable to other deep learning and computer vision challenges, such as face recognition, optical character recognition, and self-driving vehicles.

Object detection algorithms
Keypoint detection algorithms
Transfer Learning
Fully convolutional networks
Learning and using the main building blocks and layers of deep learning models
Training deep learning models on images
Deploying deep learning models

features

Self-paced: You choose the schedule and decide how much time to invest as you build your project.
Project roadmap: Each project is divided into several achievable steps.
Peer support: Chat with other participants within the liveProject platform.
Compare with others: For each step, compare your deliverable to the solutions by the author and other participants.
Book and video resources: Excerpts from Manning books and videos are included, as well as references to other resources.