Pose Estimation and Matching

Team Members

Matthew Avallone, Siddharth Choudhary, and Kshitija Patel

Overview

The goal of this project is to perform human pose estimation and matching. Given two images, a reference image and a test image, human poses would be detected using keypoints and compared to determine the similarity between them. These key points can be plotted as a 2D stick-figure. The reference image would be overlayed on the test image to find the similarity between the poses which are depicted in both of the images.

For further details, please refer to the project report.

Libraries Used

• OpenCV
• Numpy
• Matplotlib
• Pandas

Links to the prototxt and pre-trained model of the MPII Pose dataset:

• Prototxt

• Caffe Model