2D-Human-Pose-Estimation (CSC 542 Neural Networks Project - Group G33)

Group Project Members

• Aishwarya Seth
• Ishan Bhatt
• Akhil Bommadevara

Introduction

The system aims to evaluate human poses from a given set of 2D RGB images. It detects poses based on the keypoints detected from the heatmaps of images.

Dataset

The dataset used for training is derived from the MPII Human Pose Dataset, with the threshold of number of keypoints visible set to 12 in each image.

Methodology

We have used 3 hourglass models stacked one after the other for human pose estimation; the architecture being alternatively known as the stacked hourglass model.

Files

• 'Code' directory contains all the files used for processing the data and preparing the dataset. It also contains the model.
• 'Doc' directory contains all the documents related to the project - Proposal, Report, Presentation, Poster and some of the research papers referred.
• 'anno_list.csv' contains the annotations of keypoints in all images.