No abstract

Recent advances in mobile virtual reality (VR) devices have paved the way for various VR applications in education. This paper presents a novel authoring framework for mobile VR contents and a play-based learning model for marine biology education. For interactive and immersive mobile VR contents, we develop a multilayer 360° VR representation with image-based interactions such as mesh deformation and water simulation, which enable users to realistically interact with 360° panoramic contents without consuming excessive computational resources. On the basis of this representation, we design and implement play-based learning scenarios to increase the interactivity and immersion of users. Then, we verify the effectiveness of our educational scenarios using a user study in terms of user-created VR contents, interactivity, and immersion. The results show that more experienced elements in VR contents improve the immersion of users and make them more actively involved.

No abstract

To cope with the ever-increasing computational demand of the DNN execution, recent neural architecture search (NAS) algorithms consider hardware cost metrics into account, such as GPU latency. To further pursue a fast, efficient execution, DNN-specialized hardware accelerators are being designed for multiple purposes, which far-exceeds the efficiency of the GPUs. However, those hardware-related metrics have been proven to exhibit non-linear relationships with the network architectures. Therefore it became a chicken-and-egg problem to optimize the network against the accelerator, or to optimize the accelerator against the network. In such circumstances, this work presents DANCE, a differentiable approach towards the co-exploration of the hardware accelerator and network architecture design. At the heart of DANCE is a differentiable evaluator network. By modeling the hardware evaluation software with a neural network, the relation between the accelerator architecture and the hardware metrics becomes differentiable, allowing the search to be performed with backpropagation. Compared to the naive existing approaches, our method performs co-exploration in a significantly shorter time, while achieving superior accuracy and hardware cost metrics.

Microneedles (MNs) are a new system of effective drug delivery that create micron-sized pathways to the epidermis or upper dermis regions of the skin. In this study, we developed coated-type microneedles for direct hispidin delivery to the skin. Hispidin is a well-known plant-derived antioxidant component showing antitumor, anti-inflammatory, antiallergic, antiangiogenic, antioxidant, hypoglycemic, hypolipidemic, and immunomodulatory activities. Polymeric blends of polylactic acid (PLA) and polycaprolactone (PCL) were casted as MNs to enhance skin permeability. PLA/PCL MNs exhibited the highest strength of 51.26 MPa with a width of ~200 ųm. Hispidin was directly coated onto the MNs with PLA/PCL blends to form delivery layers. Compared to the hispidin-only delivery layer, skin permeability of hispidin increased by over 50% when using agarose gel in in vitro tests. In a dose-dependent manner, hispidin coated on PLA/PCL MNs also showed a brightening effect, as well as anti-inflammatory activity at the gene and protein level in skin cell culture experiments. It also demonstrated antimicrobial activity, and showed no cytotoxicity to skin cells. These results suggest that the PLA/PCL MN system with hispidin may have great potential as a prototype platform for various drug delivery systems, allowing the development of more effective subcutaneous delivery of vaccines, oligonucleotides, insulin, and many other cosmetic applications.