Most of the existing 3D human pose estimation approaches mainly focus on predicting 3D positional relationships between the root joint and other human joints (local motion) instead of the overall trajectory of the human body (global motion). Despite the great progress achieved by these approaches, they are not robust to global motion, and lack the ability to accurately predict local motion with a small movement range. To alleviate these two problems, we propose a relative information encoding method that yields positional and temporal enhanced representations. Firstly, we encode positional information by utilizing relative coordinates of 2D poses to enhance the consistency between the input and output distribution. The same posture with different absolute 2D positions can be mapped to a common representation. It is beneficial to resist the interference of global motion on the prediction results. Second, we encode temporal information by establishing the connection between the current pose and other poses of the same person within a period of time. More attention will be paid to the movement changes before and after the current pose, resulting in better prediction performance on local motion with a small movement range. The ablation studies validate the effectiveness of the proposed relative information encoding method. Besides, we introduce a multi-stage optimization method to the whole framework to further exploit the positional and temporal enhanced representations. Our method outperforms state-of-the-art methods on two public datasets. Code is available at https://github.com/paTRICK-swk/Pose3D-RIE.

Human tumors harbor a plethora of microbiota. It has been shown that the composition and diversity of intratumor microbiome are significantly associated with the survival of patients with pancreatic ductal adenocarcinoma (PDAC). However, the association in Chinese patients as well as the effect of different microorganisms on inhibiting tumor growth are unclear. In this study, we collected tumor samples resected from long-term and short-term PDAC survivors and performed 16S rRNA amplicon sequencing. We found that the microbiome in samples with different survival time were significantly different, and the differential bacterial composition was associated with the metabolic pathways in the tumor microenvironment. Furthermore, administration of Megasphaera, one of the differential bacteria, induced a better tumor growth inhibition effect when combined with the immune checkpoint inhibitor anti-programmed cell death-1 (anti-PD-1) treatment in mice bearing 4T1 tumor. These results indicate that specific intratumor microbiome can enhance the anti-tumor effect in the host, laying a foundation for further clarifying the underlying detailed mechanism.

3D human hand pose estimation from visual data has received an increasing amount of attention, and the availability of low-cost depth cameras gives a great impetus to the development of this field. Nearly all recent hand pose estimation methods are oriented towards unified evaluation criteria defined by popular public benchmark datasets: the ultimate goal is to reduce the estimation error. However, the fact is that there exists a gap between human hand pose estimation and its applications. It is unclear how to recover global and local degrees of freedom from a set of structural hand joints, which is a necessary condition to apply human hand pose estimation to teleoperation, i.e., mapping estimated human hand poses at the master side to robotic hand poses at the slave side. Conventional teleoperation systems are implemented with the aid of a data glove or essentially built on gesture recognition. These solutions are inferior to vision-based hand pose estimation in offering an easy-to-use and natural human-robot interaction interface. In this paper, we propose three methods to teleoperate robotic hands by 3D vision-based human hand pose estimation. The feasibility of the three methods is tested in a simulated environment.

No abstract

No abstract