This paper presents HomE, a homogeneous ensemble framework that leverages clustering and LLM-driven sampling to partition gesture classes, enabling expert learners to improve accuracy and robustness in dynamic hand gesture recognition.
We study the problem of robot navigation in dense and interactive crowds with environmental constraints such as corridors and furniture. Previous methods fail to consider all types of interactions among agents and obstacles, leading to unsafe and inefficient robot paths. In this article, we leverage a graph-based representation of crowded and constrained scenarios and propose a structured framework to learn robot navigation policies with deep reinforcement learning. We first split the representations of different components in the environment, and propose a heterogeneous spatio-temporal graph to model distinct interactions among humans, robots, and obstacles. Based on the heterogeneous st-graph, we propose HEIGHT, a novel navigation policy network architecture with different components to capture heterogeneous interactions among entities through space and time. HEIGHT utilizes attention mechanisms to prioritize important interactions and a recurrent network to track changes in the dynamic scene over time, encouraging the robot to avoid collisions adaptively. Through extensive simulation and real-world experiments, we demonstrate that HEIGHT outperforms state-of-the-art baselines in terms of success and efficiency in challenging navigation scenarios. Furthermore, we demonstrate that our pipeline achieves better zero-shot generalization capability than previous works when the densities of humans and obstacles change.
The demand for advanced video generation techniques has surged with the increasing application of artificial intelligence in multimedia and human-computer interaction systems. In this context, video generation models that can produce high-quality outputs with specific class or textual content are gaining prominence. In this research, we propose novel structured models for generating in-car hand gesture videos based on specific classes and explore the capability of generating new and diverse gesture videos, contributing to the growing field of automated video generation.
SmartNICs have become increasingly used to support intensive data center operations in the cloud computing industry. Today, SmartNICs are used to perform high-data-rate as well as computationally heavy tasks such as load balancing, DNS filtering, and implementing firewalls for intrusion detection. However, SmartNICs have not been explored to support use cases of resiliency that are often seen in safety-critical industries such as the Power Grid and Industrial Automation. This research aims to evaluate the feasibility of using SmartNICs to support safety-critical applications, such as fault-tolerant routing and cryptographical operations.