Recently, there has been significant progress in 2D pose estimation. However, accurately localizing limb keypoints and occluded keypoints is still challenging. To tackle these difficulties, prior in human body structure has been leveraged in previous studies. One approach involves localizing a challenging keypoint by utilizing its neighbor keypoint. A previous study successfully employed neighbor-joint spatial relation (SR), which transfers features from a neighbor keypoint to the target keypoint being predicted. Building upon this idea, our work extends the keypoint relation-based method by incorporating another level of keypoint relation, namely channel-wise feature relation. This additional feature relation (FR) module assists in selecting more suitable neighbor keypoint feature channels and enhances the effectiveness of SR. By combining FR and SR, we develop a simple and intuitive bi-level keypoint relation module that can be trained end-to-end with existing methods. Through comprehensive experimental results and ablation studies, we demonstrate the effectiveness of our approach.

For large language model (LLM) acceleration, FPGAs face two challenges: insufficient peak computing performance and unacceptable accuracy loss of model compression. This paper proposes FMC-LLM to enable FPGAs for efficient batched decoding of 70B+ LLMs.

The Cellular Potts Model (CPM) is a powerful computational framework for simulating collective cell behavior, including morphogenesis, cell migration, and tissue dynamics. However, conventional CPM implementations rely on a sequential Modified Metropolis Algorithm (MMA), making large-scale simulations computationally expensive. In this work, we present a parallel boundary-based algorithm that accelerates CPM simulations by restricting copy attempts to boundary sites (i.e., lattice sites located at cell-cell or cell-substrate interfaces). Our algorithm combines local energy evaluation with a thread-safe parallelization strategy in which multiple threads independently assess copy attempts, while a central core thread applies accepted updates sequentially to ensure consistency. Benchmark results demonstrate a significant reduction in computation time compared to conventional implementations, enabling efficient simulation of larger and more complex cellular systems.

A 23–29 GHz power amplifier (PA) based on a stacked field-effect transistor (FET) topology is proposed. To achieve higher output power and gain, the output stage employs a triple-stacked FET design. To improve stacking efficiency, neutralization capacitors are added between the differential common-source (CS) amplifiers to compensate for phase mismatches at the stack nodes. Round-table and vertical gate repetition techniques are applied to achieve a compact layout and enhance the PA’s power density. The proposed PA is fabricated using the 22nm FD-SOI process. Measurement results show that the proposed PA delivers a peak saturated power (Psat) of 19.6 dBm, with a core area of only 0.047 mm2, achieving a power density of 1936.2 mW/mm2.

In recent years, sketch-based 3D shape retrieval has attracted growing attention. While many previous studies have focused on cross-modal matching between hand-drawn sketches and 3D shapes, the critical issue of how to handle low-quality and noisy samples in sketch data has been largely neglected. This paper presents an uncertainty-aware cross-modal transfer network (UACTN) that addresses this issue. UACTN decouples the representation learning of sketches and 3D shapes into two separate tasks: classification-based sketch uncertainty learning and 3D shape feature transfer. We first introduce an end-to-end classification-based approach that simultaneously learns sketch features and uncertainty, allowing uncertainty to prevent overfitting noisy sketches by assigning different levels of importance to clean and noisy sketches. Then, 3D shape features are mapped into the pre-learned sketch embedding space for feature alignment. Extensive experiments and ablation studies on two benchmarks demonstrate the superiority of our proposed method compared to state-of-the-art methods.

Creating a large-scale functional structure from a group of bio-nanomachines is key for engineering applications of molecular communication. This paper aims to create a large-scale functional structure from a group of bio-nanomachines, and proposes a collective rotational motion model of bio-nanomachines. In the proposed model, a group of bio-nanomachines forms a cluster that continues to rotate. The proposed model is based on the idea that spinning objects are stable against perturbations. In developing the model, we draw inspiration from biological pattern formation: biological entities interact chemically and physically to form a functional structure. Accordingly, we develop a collective rotational motion model based on chemical and physical interactions between bio-nanomachines. Through modeling and simulations, we gain insight into design and engineering of rotating clusters of bio-nanomachines. This paper demonstrates the importance of physical interactions in creating a system-level functionality from a group of bio-nanomachines, giving rise to new challenges and opportunities in molecular communication research.

In skeleton-based action recognition, graph convolutions to model human action dynamics have been widely implemented and achieved remarkable results. Among these convolutions, channel-wise adaptive graph convolution shows outstanding performance. However, this method focuses too much on capturing correlation between joints within each channel and lacks the capability of learning structural features, which are generally hidden in geometric property of the skeleton on spatial domain. Our proposed method (SA-GCN) introduces symmetry trajectory attention module to measure the relation between left and right part of body and part relation attention module for exploration of the attention on general relation of each part. Both modules are intended to make full use of structural features in skeleton, further strengthening advantages of graph convolution. Experiments on three datasets (NW-UCLA, NTU-RGB+D and NTU-RGB+D 120) demonstrate state-of-the-art performance of our model, especially on joint modality.