MobileHumanPose: Toward Real-Time 3D Human Pose Estimation in Mobile Devices
Published in IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2021
Recommended citation: Sangbum Choi, Seokeon Choi, and Changick Kim. (2021). "MobileHumanPose: Toward Real-Time 3D Human Pose Estimation in Mobile Devices." IEEE/CVF CVPR Workshops, 2328-2338. https://openaccess.thecvf.com/content/CVPR2021W/MAI/papers/Choi_MobileHumanPose_Toward_Real-Time_3D_Human_Pose_Estimation_in_Mobile_Devices_CVPRW_2021_paper.pdf
MobileHumanPose is a mobile-friendly 3D human pose estimation model built for real-time inference from a single RGB image. The model uses a modified MobileNetV2 backbone, parametric activation, and U-Net-inspired skip concatenation to reduce model size and improve CPU inference speed on mobile devices.