Occluded Human Mesh Recovery

Abstract

Figure 1: Left: Our proposed capture setup consists of multiple egocentric cameras from wearable glasses and stationary secondary cameras. This setup is flexible and mobile, allowing us to generate high-quality multi-human 3D annotations for diverse in-the-wild settings. Center: Multiple synchronized egocentric views while playing soccer. Right: Synchronized secondary views (cropped) from the stationary cameras. All cameras are spatiotemporally localized in the world coordinate.

Medicine-Engineering Interdisciplinary Research Based on Bibliometric Analysis: A Case Study on Medicine-Engineering Institutional Cooperation of Shanghai Jiao Tong University

“…Our results indicate that DIMR, achieves comparable performance when compared to state-of-the-art methods. Furthermore, we observe that Methods 3DPW PA-MPJPE↓ MPJPE↓ PVE↓ General DecoMR [49] 61.7 --SPIN [27] 59.2 96.9 116.4 PyMAF [43] 58.9 92.8 110.1 HMR-EFT [20] 52.2 85.1 98.7 METRO [32] 47.9 77.1 88.2 Occludsion OCHMR [22] 58.3 89.7 107.1 Pose2UV [16] 57.1 --ROMP [44] 53.3 85.5 103.1 PARE [25] 50.9 82 97.9 DIMR (Ours) 52.9 87 106.6 DIMR (GT IUV) 47.8 73.0 99.5…”

“…Evaluation metrics. Our evaluation metrics on the 3D datasets include mean per joint position error (MPJPE), [22,46], 3DPW-OC [25,46] and 3DOH [52]. For a fair comparison, we train our models without 3DPW datasets when evaluate on both of the 3DPW subset.…”

“…Jiang et al [36] utilize an interpenetration loss to regular multiperson location based on instance segmentation. 3DCrowdNet [22] proposes to distinguish the target's image feature with the occluded human with the guide of a 2D joint map. Compared to the sparse representation of a predicted human, our model utilizes a dense correspondence map to achieve accurate and dense image feature separation.…”

“…Two-stage methods designed to predict one person at a time tend to focus on separating and identifying the true human to predict. OCHMR [22] utilizes a ROMPliked center heat map. 3DCrowdNet [7] uses a keypoint heatmap.…”

An Outlook into the Future of Egocentric Vision

“…They argue that accurate motion prediction depends on understanding human intentions, which can be studied using gaze in the egocentric setting. Additionally, the EgoHumans benchmark by Khirodkar et al (2023) captures multiple subjects in realistic outdoor environments from multiple egocentric viewpoints, serving as a valuable resource for multi-view multi-human analysis.…”