Specific Pose Detection Through Efficient Neural Network

Human pose detection is a process of detecting human body points in images or videos. For the past few years, it has gained the attention of the computer vision community. Some of the achievements such as single pose detection and top-down network were efficient and reliable but suffered some problems. This technology has a very vast scope and is applicable in areas such as medical industry, defense industry, public safety, personal monitoring, and mentoring but due to the limitation of technology and algorithm, implementation has become a big hassle. In this work, we bring a real-time network-based pose estimation system that utilizes CPU processing and generates the normalized body points concerning the frame. It demonstrates how these points can be used to calculate the different body part angles as well as their lengths and by using these calculations, classifying the pose into a category to detect and output the result in the desired format. Our project outperformed the current system in terms of efficiency, cost, and processing speed.