Evaluation of Deep Learning based Pose Estimation for Sign Language

Srujana Gattupalli, Amir Ghaderi, Vassilis Athitsos

Published 2016-02-29Version 1

Human body pose estimation and hand detection being the prerequisites for sign language recognition(SLR), are both crucial and challenging tasks in Computer Vision and Machine Learning. There are many algorithms to accomplish these tasks for which the performance measures need to be evaluated for body posture recognition on a sign language dataset, that would serve as a baseline to provide important non-manual features for SLR. In this paper, we propose a dataset for human pose estimation for SLR domain. On the other hand, deep learning is on the edge of the computer science and obtains the state-of-the-art results in almost every area of Computer Vision. Our main contribution is to evaluate performance of deep learning based pose estimation methods by performing user-independent experiments on our dataset. We also perform transfer learning on these methods for which the results show huge improvement and demonstrate that transfer learning can help improvement on pose estimation performance of a method through the transferred knowledge from another trained model. The dataset and results from these methods can create a good baseline for future works and help gain significant amount of information beneficial for SLR.