Vision for Human Pose Estimation
Moving Objects Segmentation and Background Estimation
January 25, 2024AI for Graph Data Analysis
January 25, 2024
In this vertical we advance computer vision techniques to precisely capture and interpret human poses, fostering applications in health monitoring, sports analysis, and human-machine interaction.
a. Quantification of Occlusion Handling Capability of a 3D Human Pose Estimation Framework
In the domain of 3D human pose estimation from monocular images, the
challenge of handling occlusion remains less focused. This work introduces
occlusion-guided 3D human pose estimation framework that use 2D skeletons
with missing joints as input. Occlusion guidance is incorporated to provide
additional information about a joint absence or presence, enhancing accuracy. We
use temporal information to improve estimation when joints are missing.
Related Publications
b. A Boosting Framework for Human Posture Recognition
In many surveillance applications, automatic human posture recognition is crucial
for monitoring various activities. To overcome challenges like occlusion,
background clutter, and illumination variations, we combine spatiotemporal
features such as aspect ratios, shape descriptors, geometric centroids, ellipse axes
ratio, silhouette angles, and silhouette speed. We also leverage Radon Transform
for enhanced shape-based analysis.
Related Publications
c. Walk Like Me: Video-to-Video Action Transfer
In this study, we explore different methods for transferring human actions from a
source to a target video with improved human motion smoothness and better
image quality. The project focuses on video-to-video action transfer GAN based
algorithms that have achieved better image quality by employing a cascaded
sequence of action transfer blocks with multi-resolution structure similarity (MR-
SSIM) loss.
Related Publications
d. Human Fall Detection
Fall-induced damages are serious incidences for aged as well as young persons. A
real-time automatic and accurate fall detection system can play a vital role in
timely medication care which will ultimately help to decrease the damages and
complications. We explore fast and more accurate real-time systems that can
detect people falling in videos captured by surveillance cameras. Novel temporal
and spatial variance-based features are proposed which comprise the
discriminatory motion, geometric orientation, and location of a falling person.
Related Publications