Paper ID | D-3-3.8 |
Paper Title |
Human Pose Estimation Using Skeletal Heatmaps |
Authors |
Jinyoung Jun, Jae-Han Lee, Chang-Su Kim, Korea University, Korea (South) |
Session |
D-3-3: Image and video processing based on deep learning |
Time | Thursday, 10 December, 17:30 - 19:30 |
Presentation Time: | Thursday, 10 December, 19:15 - 19:30 Check your Time Zone |
|
All times are in New Zealand Time (UTC +13) |
Topic |
Image, Video, and Multimedia (IVM): Special Session: Image and video processing based on deep learning |
Abstract |
We propose a novel skeletal attention module to generate keypoint heatmaps, which exploits skeletal, as well as overall body structure, information for human pose estimation. We first add augmenting convolutional layers to an existing deep neural network in order to yield skeletal heatmaps. These skeletal heatmaps emphasize keypoint relations connected either physically or virtually. By combining the skeletal heatmaps, we generate body attention maps for upper-body, lower-body, and full-body. Then, the skeletal heatmaps and the body attention maps are employed to estimate the heatmap for each keypoint. Finally, we perform weighted inference on the output heatmaps for more precise estimates. Experimental results demonstrate that the proposed algorithm enhances performance on two datasets for human pose estimation. |