Golf posture detection by following along, Event Detection using Human Pose

Everything You Need to Know About 3D Human Pose Event Detection! 🚩

Light with papers, fun with practice
3D Human Pose Event Detection!

How can you detect someone who has collapsed in an empty elevator? This technology originates from the field of Video Understanding, a computer vision discipline that recognizes specific actions in videos and images. In this lecture, we'll learn about Event Detection, a field within Video Understanding that detects frames of specific actions, events, and scenes in videos. You can do this without thick books or boring data!

I entered graduate school with the goal of becoming an AI expert, but I had no research experience, no papers to write, and no reading. All I could do was follow along with blog posts. As a result, I struggled quite a bit during my first year of graduate school. Based on those experiences, I designed this course. The appeal of this course lies in its ability to provide students with insight and creative thinking. While the detailed experimental procedures are omitted, I will provide guidance on how research is conducted.

This course aims to briefly review relevant papers, run the paper's implementation code, and improve performance over existing methods . We'll examine the paper's network implementation code, and then train, evaluate, test, and even design the network. I believe learning new areas isn't about studying and practicing, but rather about practicing. Let's give it a try!

Learn this 📚

We will study the field of detecting and recognizing specific events or scenes in video, called Event Spotting or Event Detection.

Among them, we extract eight key motion scenes from universal and globally popular golf swing videos. Because golf swings are characterized by distinct movements and posture is crucial in the sport, many golfers want to analyze their own posture. To achieve this, we will develop an artificial intelligence model using event detection.

If we extract eight action scenes from a golf video, we can create an artificial intelligence model that captures eight postures as shown below.

In the existing method, we train the network by providing a new input, Human Pose.

Introducing the Knowledge Sharer ✒️

Knowledge sharer who created this course
Introducing “Copycat” Kim Hyun-woo.

The follower who created this course is🥽

I will provide you with research and project experience based on my knowledge of computer science and artificial intelligence, experience with numerous deep learning/machine learning projects, competition awards, and graduate school research experience .

• Graduated with a master's degree in artificial intelligence from Korea University
• Oral presentation at an international conference
• ACCV, first author of a paper on pattern recognition published

Hello, I am Kim Hyeon-woo, a copycat.

While completing my Master's degree in Artificial Intelligence at Korea University in just two years, I published a paper on 3D Human Pose Estimation at ACCV2022 (Asian Conference on Computer Vision), the world's 19th-ranked conference in the field of artificial intelligence and computer vision, and it was also selected for an oral presentation. Later, I refined my original paper and published it as first author in Pattern Recognition, the world's 6th-ranked journal.

I created this course after seeing many people struggle to learn or even give up due to the lack of accessibility to certain fields of artificial intelligence. It captures the lessons learned and the trials and errors I've faced over the past two years.

I hope that through this lecture, you will gain interest in the field of artificial intelligence and gain a new sense of accomplishment :)

I recommend this to these people 🙆‍♀️

Those who want to go to graduate school in the field of artificial intelligence but have no research or thesis experience

For those who want to quickly get a taste of a field of artificial intelligence

For those who are curious about how artificial intelligence research is done

College students who need novel ideas for contests and projects

Expected Questions Q&A 💬

Things to note before taking the course 📢

Practice environment

• Operating System and Version (OS): Ubuntu
• Tools used: Visual Studio Code, Anaconda
• PC specifications: Any GPU with an RTX 10 series or higher is supported.

Learning Materials

• Completed final code folder zip file

Player Knowledge and Precautions

• Anyone who has experience with Python and is familiar with Visual Studio and Anaconda
• Those who installed Ubuntu