苏卓.pdf

1、Human Motion Capture and Avatar Creation for XRScenarios using Sparse ObservationsZhuo Su 苏 卓1.BackgroundAvatar Creation3D communicationConnecting people through lifelike 3D reconstruction for enhanced interactionMotion Capture3D GamingEnhancing gaming,social interactions,and fitness for immersive e

2、xperiences-XR Applications-Sparse ObservationsHMD and hand controllersTracking signals3DoF of motion trackersPICO Motion TrackerApple Vision Pro PersonaMeta Codec AvatarEgo-centric view imagesFront view image2.Motion Capture using Sparse SensorsEnvPoserCVPR25+3D environment from VR deviceEMHI(MEPose

3、r)AAAI25+egocentric cameras of HMDAvatarJLMICCV236DoF of HMD and hand controllers*Explore these research papers at https:/suzhuo.github.io/HMD-PoserCVPR24+scalable IMUs on legs or pelvisOHTACVPR24One-shot hand avatar creationHumanSplatNeurIPS24Single-Image full-body Gaussian SplattingHeadGap3DV25Few

4、-shot head avatar creationSEGAarXiv25One-shot head avatar creation3.Avatar Creation using One/Few Image(s)*Explore these research papers at https:/suzhuo.github.io/4.Conclusion&Future Work-Real-World Robustness:Expand datasets,optimize generalizable algorithm,and enhance real-time performance in rea

5、l-world scenarios-Integration with Video Foundation Models:Further Leverage large generative models to enhance motion realism and avatar fidelity-Seamless Motion-Avatar Integration:Tightly integrate motion capture and avatar creation to enable real-time,full-body XR experiences-Conclusion-Future Wor

6、kOver the past year,my research focused on addressing the challenges of sparse observations in XR scenarios,yielding significant achievements:-Motion Capture:Enabled high-accuracy motion capture under sparse sensor configurations-Avatar Cre