Convolutional neural networks (CNNs) require numerous computations and external memory accesses. Frequent accesses to off-chip memory cause slow processing and large power dissipation. For real-time ...object detection with high throughput and power efficiency, this paper presents a Tera-OPS streaming hardware accelerator implementing a you-only-look-once (YOLO) CNN. The parameters of the YOLO CNN are retrained and quantized with the PASCAL VOC data set using binary weight and flexible low-bit activation. The binary weight enables storing the entire network model in block RAMs of a field-programmable gate array (FPGA) to reduce off-chip accesses aggressively and, thereby, achieve significant performance enhancement. In the proposed design, all convolutional layers are fully pipelined for enhanced hardware utilization. The input image is delivered to the accelerator line-by-line. Similarly, the output from the previous layer is transmitted to the next layer line-by-line. The intermediate data are fully reused across layers, thereby eliminating external memory accesses. The decreased dynamic random access memory (DRAM) accesses reduce DRAM power consumption. Furthermore, as the convolutional layers are fully parameterized, it is easy to scale up the network. In this streaming design, each convolution layer is mapped to a dedicated hardware block. Therefore, it outperforms the "one-size-fits-all" designs in both performance and power efficiency. This CNN implemented using VC707 FPGA achieves a throughput of 1.877 tera operations per second (TOPS) at 200 MHz with batch processing while consuming 18.29 W of on-chip power, which shows the best power efficiency compared with the previous research. As for object detection accuracy, it achieves a mean average precision (mAP) of 64.16% for the PASCAL VOC 2007 data set that is only 2.63% lower than the mAP of the same YOLO network with full precision.
Early diagnosis of the highly pathogenic H5N1 avian influenza virus (AIV) is significant for preventing and controlling a global pandemic. However, there is no existing electrical biosensor for ...detecting biomarkers for AIV in clinically relevant samples such as chicken serum. Herein, we report the first use of an aptamer-functionalized field-effect transistor (FET) as a label-free sensor for AIV detection in chicken serum. A DNA aptamer is employed as a sensitive and selective receptor for hemagglutinin (HA) protein, which is a biomarker for AIVs. This aptamer is immobilized on a gold microelectrode that is connected to the gate of a reusable FET transducer. The specific binding of the target protein results in a change in the surface potential, which generates a signal response of the FET transducer. We hypothesize that a conformational change in the DNA aptamer upon specific binding of HA protein may alter the surface potential. The signal of the aptamer-based FET biosensor increased linearly with the increase in the logarithm of HA protein concentration in a dynamic range of 10 pM to 10 nM with a detection limit of 5.9 pM. The selectivity of the biosensor for HA protein was confirmed by employing relevant interfering proteins. The proposed biosensor was successfully applied to the selective detection of HA protein in a chicken serum sample. Owing to its simple and low-cost architecture, portability, and sensitivity, the aptamer-based FET biosensor has potential as a point-of-care diagnosis of H5N1 AIVs in clinical samples.
The use of object detection algorithms is becoming increasingly important in autonomous vehicles, and object detection at high accuracy and a fast inference speed is essential for safe autonomous ...driving. A false positive (FP) from a false localization during autonomous driving can lead to fatal accidents and hinder safe and efficient driving. Therefore, a detection algorithm that can cope with mislocalizations is required in autonomous driving applications. This paper proposes a method for improving the detection accuracy while supporting a real-time operation by modeling the bounding box (bbox) of YOLOv3, which is the most representative of one-stage detectors, with a Gaussian parameter and redesigning the loss function. In addition, this paper proposes a method for predicting the localization uncertainty that indicates the reliability of bbox. By using the predicted localization uncertainty during the detection process, the proposed schemes can significantly reduce the FP and increase the true positive (TP), thereby improving the accuracy. Compared to a conventional YOLOv3, the proposed algorithm, Gaussian YOLOv3, improves the mean average precision (mAP) by 3.09 and 3.5 on the KITTI and Berkeley deep drive (BDD) datasets, respectively. Nevertheless, the proposed algorithm is capable of real-time detection at faster than 42 frames per second (fps) and shows a higher accuracy than previous approaches with a similar fps. Therefore, the proposed algorithm is the most suitable for autonomous driving applications.
Here, a novel, versatile synthetic strategy to fabricate a yolk–shell structured material that can encapsulate virtually any functional noble metal or metal oxide nanocatalysts of any morphology in a ...free suspension fashion is reported. This strategy also enables encapsulation of more than one type of nanoparticle inside a single shell, including paramagnetic iron oxide used for magnetic separation. The mesoporous organosilica shell provides efficient mass transfer of small target molecules, while serving as a size exclusion barrier for larger interfering molecules. Major structural and functional advantages of this material design are demonstrated by performing three proof‐of‐concept applications. First, effective encapsulation of plasmonic gold nanospheres for localized photothermal heating and heat‐driven reaction inside the shell is shown. Second, hydrogenation catalysis is demonstrated under spatial confinement driven by palladium nanocubes. Finally, the surface‐enhanced Raman spectroscopic detection of model pollutant by gold nanorods is presented for highly sensitive environmental sensing with size exclusion.
A newly developed yolk–shell structure encapsulates functional nanoparticles of complex morphology to enable efficient localized photothermal heating, hydrogenation catalysis under spatial confinement, and ultrasensitive environmental sensing based on surface‐enhanced Raman spectroscopy. The mesoporous organosilica shell facilitates selective transport of target reactant and analyte from bulk phase to the inner reaction compartment, while rejecting interfering species via size exclusion.
Convolutional neural networks (CNNs) require both intensive computation and frequent memory access, which lead to a low processing speed and large power dissipation. Although the characteristics of ...the different layers in a CNN are frequently quite different, previous hardware designs have employed common optimization schemes for them. This paper proposes a layer-specific design that employs different organizations that are optimized for the different layers. The proposed design employs two layer-specific optimizations: layer-specific mixed data flow and layer-specific mixed precision. The mixed data flow aims to minimize the off-chip access while demanding a minimal on-chip memory (BRAM) resource of an FPGA device. The mixed precision quantization is to achieve both a lossless accuracy and an aggressive model compression, thereby further reducing the off-chip access. A Bayesian optimization approach is used to select the best sparsity for each layer, achieving the best trade-off between the accuracy and compression. This mixing scheme allows the entire network model to be stored in BRAMs of the FPGA to aggressively reduce the off-chip access, and thereby achieves a significant performance enhancement. The model size is reduced by 22.66-28.93 times compared to that in a full-precision network with a negligible degradation of accuracy on VOC, COCO, and ImageNet datasets. Furthermore, the combination of mixed dataflow and mixed precision significantly outperforms the previous works in terms of both throughput, off-chip access, and on-chip memory requirement.
Methylglyoxal (MGO), which is produced as a byproduct of glucose metabolism, is the leading to diabetic cardiovascular complications. Salvia miltiorrhiza Bunge (Lamiaceae) has been reported as a ...potential plant to control diabetes and cardiovascular disease. However, no report exists on the effect of Salvia miltiorrhiza Bunge extract (SME) on MGO-induced glucotoxicity in human umbilical vein endothelial cells (HUVECs). We demonstrated the protective effects of SME (1, 5, and 10 µg/mL) and its components against MGO-induced endothelial dysfunction in HUVECs. Cytotoxicity was evaluated using the several in vitro experiments. Additionally, the protein expression of receptor of advanced glycation end-products (RAGE), mitogen-activated protein kinase (MAPK) pathway and glyoxalase system were measured. Then, the inhibitory effects of SME and its main components on MGO-induced oxidative stress, radical scavenging, formation of MGO-derived advanced glycation end products (AGEs), and MGO-AGEs crosslinking were evaluated. SME (10 µg/mL) strongly prevented expressed levels of RAGE, MGO-induced apoptosis and reduced reactive oxygen species (ROS) generation in HUVECs, comparing with 1 mM aminoguanidine. Additionally, SME (5 and 10 µg/mL) reduced the expression of proteins (e.g., p-extracellular signal-regulated kinase (ERK) and p-p38) in the MAPKs pathway and upregulated the glyoxalase system in HUVECs. SME (0.5–10 mg/mL), dihydrotanshinone (0.4 mM), and rosmarinic acid (0.4 mM) prevented MGO-AGEs formation and broke the MGO-AGE crosslinking. These results show that S. miltiorrhiza has protective effects against MGO-induced glucotoxicity by regulating the proteins involved in apoptosis, glyoxalase system and antioxidant activity. We expect that S. miltiorrhiza is a potential natural resource for the treatment of MGO-induced vascular endothelial dysfunction.
The purpose of this study was to compare the fixation stability of proximal fragments and the mechanical characteristics in proximal femur models of basicervical femoral neck fracture fixed by the ...femoral neck system (FNS) versus the dynamic hip screw. The mean axial stiffness was 234 ± 35 N/mm in the FNS group and 253 ± 42 N/mm in the DHS group, showing no significant difference (p = 0.654). Mean values for x-axis rotation, y-axis rotation, and z-axis rotation after cycle load were 2.2 ± 0.5°, 6.5 ± 1.5°, and 2.5 ± 0.6°, respectively, in the FNS group and 2.5 ± 0.7°, 5.8 ± 2.1°, and 2.2 ± 0.9°, respectively, in the DHS group, showing no significant differences (p = 0.324, p = 0.245, and p = 0.312, respectively). The mean values of cranial and axial migration of screws within the femoral head were 1.5 ± 0.3 and 2.1 ± 0.2 mm, respectively, in the FNS group and 1.2 ± 0.3 and 2.4 ± 0.3 mm, respectively, in the DHS group, showing no significant differences (p = 0.425 and p = 0.625, respectively). The average failure load at vertical load was 1342 ± 201 N in the FNS group and 1450 ± 196 N in the DHS group, showing no significant difference (p = 0.452). FNS fixation might provide biomechanical stability comparable to that of DHS for treating displaced basicervical femoral neck fractures in young adults.
SnSe emerges as a new class of thermoelectric materials since the recent discovery of an ultrahigh thermoelectric figure of merit in its single crystals. Achieving such performance in the ...polycrystalline counterpart is still challenging and requires fundamental understandings of its electrical and thermal transport properties as well as structural chemistry. Here we demonstrate a new strategy of improving conversion efficiency of bulk polycrystalline SnSe thermoelectrics. We show that PbSe alloying decreases the transition temperature between Pnma and Cmcm phases and thereby can serve as a means of controlling its onset temperature. Along with 1% Na doping, delicate control of the alloying fraction markedly enhances electrical conductivity by earlier initiation of bipolar conduction while reducing lattice thermal conductivity by alloy and point defect scattering simultaneously. As a result, a remarkably high peak ZT of ∼1.2 at 773 K as well as average ZT of ∼0.5 from RT to 773 K is achieved for Na0.01(Sn1–x Pb x )0.99Se. Surprisingly, spherical-aberration corrected scanning transmission electron microscopic studies reveal that Na y Sn1–x Pb x Se (0 < x ≤ 0.2; y = 0, 0.01) alloys spontaneously form nanoscale particles with a typical size of ∼5–10 nm embedded inside the bulk matrix, rather than solid solutions as previously believed. This unexpected feature results in further reduction in their lattice thermal conductivity.
Due to unparalleled theoretical capacity and operation voltage, metallic Li is considered as the most attractive candidate for lithium‐ion battery anodes. However, Li metal electrodes suffer from ...uncontrolled dendrite growth and consequent interfacial instability, which result in an unacceptable level of performance in cycling stability and safety. Herein, it is reported that a marginal amount (1.5 at%) of magnesium (Mg) doping alters the surface properties of Li metal foil drastically in such a way that upon Li plating, a highly dense Li whisker layer is induced, instead of sharp dendrites, with enhanced interfacial stability and cycling performance. The effect of Mg doping is explained in terms of increased surface energy, which facilitates plating of Li onto the main surface over the existing whiskers. The present study offers a useful guideline for Li metal batteries, as it largely resolves the longstanding shortcoming of Li metal electrodes without significantly sacrificing their main advantages.
Marginal magnesium doping, (1.5 at%), alters the surface properties of Li metal foil drastically, such that a highly compact Li layer is induced upon Li plating, instead of troublesome dendrite formation, resulting in markedly improved long‐term battery performance. Density functional theory calculations capture the enhanced lithiophilicity of Li metal by magnesium doping.