Convolutional neural network (CNN) based object detection algorithms are becoming dominant in many application fields due to their superior accuracy advantage over traditional schemes. Among them, You Look Only Once (YOLO) is one of the most popular frameworks that show best trade-offs between speed and accuracy. However, intrinsic high computational workload CNN, it still challenging when targeting high-throughput processing with low cost energy consumption. In this paper, we propose a...

The efficient implementation of high-rate high-throughout low-density parity-check (LDPC) code decoding presents challenges to both fully parallel and memory-sharing partially schemes. In this paper, a new scheme, sliced message passing (SMP), is introduced. key idea slice the total set N variable-to-check messages into equal-sized p chunks, then perform check-node computation sequentially chunk by chunk. This scheme can break sequential tie between check- variable-node update stages thus...

A concept of a quadtree based nonsquare block coding structure is presented in this paper for the emerging High Efficiency Video Coding standard, which includes asymmetric motion partitioning scheme and transform (NSQT) algorithm. Nonsquare inter frame provides possibility getting more accurate prediction results by splitting one into two blocks. Contrary to traditional symmetric partitions (SMP), (AMP) are proposed improve efficiency, especially blocks with irregular object boundaries. NSQT...

A two-photon excitable small organic molecule (abbreviated as TP-NH2) with large absorption cross section and competitive fluorescence quantum yield was prepared, which emitted in the visible region upon excitation at 800 nm. Using TP-NH2 an energy donor, a resonance energy-transfer (TPE-FRET) based homogeneous immunoassay method proposed. The donor acceptor (DABS-Cl, dark quencher) were labeled to bovine serum albumin (BSA) separately, anti-BSA protein determined by employing antibody...

The influence of different synthetic pathways on the phase transition and side-chain crystallization behavior cellulose ethers was systematically investigated by a combination differential scanning calorimetry (DSC), small wide-angle X-ray scattering (SAXS, WAXS) Fourier transform infrared spectroscopy (FTIR) techniques. DSC investigation indicated that melting temperature temperature, respectively, derivatives were synthesized homogeneous reactions in solutions differed from those...

AbstractTo enhance soil management of Malus hupehensis orchards, we investigated the effects maize residue (MR) and cellulose-decomposing bacteria on microbial properties, organic fractions, growth 1-year-old potted M. Rehd. seedlings. Potted was treated with MR, inocula (MI) Bacillus cereus (MI), MR + MI, untreated (CK), collected after 30, 60, 90 d. The results showed that increased biomass carbon (MBC), MI reduced MBC. Biolog Eco-Plate™ functional diversity significantly higher at 30 d...

Advances in moving object detection have been driven by the active application of deep learning methods. However, many existing models render superior accuracy at cost high computational complexity and slow inference speed. This fact has hindered development such mobile embedded vision tasks, which need to be carried out a timely fashion on computationally limited platform. In this paper, we propose super-fast (inference speed-154 fps) lightweight (model size-1.45 MB) end-to-end 3D separable...

A design of 2-D forward and inverse integer transform processor is presented, which suitable for MPEG-4 AVC/H.264 visual profile. The comparability between the symmetry their arithmetic has been utilized in architecture. According to this design, implemented by using duplicated 1-D transform. Parallel register array are used realize transpose operation. Under 0.35um technology, logic gate count only 3524 when maximum frequency more than 120MHz.

This paper proposes a new intra coding scheme, known as short distance prediction (SDIP), for high efficiency video (HEVC) standardization work. The proposed method is based on the quadtree unit structure of HEVC. By splitting into nonsquare units and reconstruction, therefore shortening distances between predicted reference samples, accuracy can be improved when applying directional method. SDIP improves accuracy, especially high-detailed regions. approach applied in both luma chroma...

Transformers have been an indispensable staple in deep learning. However, it is challenging to realize efficient deployment for Transformer-based model due their substantial computation and memory demands. To address this issue, we present sparse Transformer accelerator on FPGA, namely STA, by exploiting N:M fine-grained structured sparsity. Our design features not only a unified computing engine capable of performing both dense matrix multiplications with high computational efficiency, but...

A design of 2-D forward and inverse integer transform processor is presented, which suitable for MPEG-4 AVC/H.264 visual profile. The comparability between the symmetry their arithmetic has been utilized in architecture. According to this design, implemented by using duplicated 1-D transform. Parallel register array are used realize transpose operation. Under 0.35um technology, logic gate count only 3524 when maximum frequency more than 120MHz.

Nitrogen (N) cycling microorganisms mediate soil nitrogen transformation processes, thereby affecting agricultural production and environment quality. However, it is not fully understood how active N-cycling microbial community in respond to long-term fertilization, as well which regulate ecosystem. Here, we collected the soils from different depths seasons at a 29-year fertilization experimental field (organic/chemical fertilizer), investigated transcriptions of functional genes their...

Generative adversarial network (GAN)-based image compression approaches reconstruct images with highly realistic quality at low bit rates. However, currently there is no published GAN-based approach that utilizes advanced GAN losses, such as the Wasserstein gradient penalty loss (WGAN-GP), to improve of reconstructed images. Meanwhile, existing deep learning-based require extra convolution layers estimate and constrain entropy during training, which makes larger may bits send information...

Abstract This study investigated the effects of forage species and sampling depths on microbial community functional diversity in soil collected between rows apple trees. Apple orchard soils' intercropping three different species, including native wild grasses, red clover (RC), ryegrass, clean tillage (CT) at varying (0–10, 10–20, 20–30 cm), was analyzed using Biolog Eco-PlateTM. The average well color development (AWCD) samples subject to ELISA indicated that carbon metabolic capacities RC...