Product posters blend striking visuals with informative text to highlight the product and capture customer attention. However, crafting appealing posters and manually optimizing them based on online performance is laborious and resource-consuming. To address this, we introduce AutoPP, an automated pipeline for product poster generation and optimization that eliminates the need for human intervention. Specifically, the generator, relying solely on basic product information, first uses a unified design module to integrate the three key elements of a poster (background, text, and layout) into a cohesive output. Then, an element rendering module encodes these elements into condition tokens, efficiently and controllably generating the product poster. Based on the generated poster, the optimizer enhances its Click-Through Rate (CTR) by leveraging online feedback. It systematically replaces elements to gather fine-grained CTR comparisons and utilizes Isolated Direct Preference Optimization (IDPO) to attribute CTR gains to isolated elements. Our work is supported by AutoPP1M, the largest dataset specifically designed for product poster generation and optimization, which contains one million high-quality posters and feedback collected from over one million users. Experiments demonstrate that AutoPP achieves state-of-the-art results in both offline and online settings.

ABSTRACT Functional near‐infrared spectroscopy (fNIRS), renowned for its high spatial resolution, shows substantial promise in brain‐computer interface (BCI) applications. However, challenges such as lengthy data acquisition processes and susceptibility to noise can limit data availability and reduce classification accuracy. To overcome these limitations, we introduce the CosineGAN‐transformer network (CGTNet), which integrates a dual discriminator GAN for generating high‐quality synthetic data with a Transformer‐based classification network. Equipped with a multi‐head self‐attention mechanism, this network excels at capturing the intricate spatiotemporal relationships inherent in high‐resolution fNIRS signals. The dual discriminator framework ensures that both the temporal and spatial aspects of the synthetic data closely resemble the original signals, thereby enhancing data diversity and fidelity. Experimental results on a publicly available fNIRS dataset, comprising 30 participants performing motor imagery tasks (right‐hand tapping, left‐hand tapping, and foot tapping), demonstrate that CGTNet achieves an accuracy of 82.67%, outperforming existing methods. Key contributions of this work include the use of multi‐head self‐attention for refined feature extraction and a dual discriminator Generative Adversarial Networks (GAN) framework that maintains data quality and consistency. These advancements significantly improve the robustness and accuracy of BCI systems, offering promising applications in neurorehabilitation and assistive technologies.

Abstract In the process of workpieces production, surface defects are prone to occur, and these defects come in a wide variety and are often intermixed, making defect detection and classification exceptionally challenging. With the development of artificial intelligence and deep learning, to tackle this problem, this paper introduces an enhanced single shot multibox detector algorithm based on MobileNetV2 for the detection of surface defects. The method utilizes MobileNetv2 as the backbone of the feature extraction network to obtain six feature layers with different detection scales from the baseline network, that is, the original 1 × 1 feature prediction layer is deleted and a 75 × 75 feature prediction layer is added, which is closer to the specific features of the defects on the surface of the workpiece. In the additional feature layer, two parallel dilated convolution structures are connected, introducing depth‐separable convolution and dilated convolution, combining skip connections and pixel‐wise addition operations. A special feature fusion structure is proposed to perform feature fusion for small, medium and large target detection layers, which effectively solves the problem of missed and false detection. Moreover, it refines candidate bounding box aspect ratios within the training set through the utilization of the K‐means clustering algorithm, ensuring a better match with real boxes. The experimental results demonstrate the effectiveness of the enhanced model, and the mean average precision value reaches 88.72%. Compared to other state‐of‐the‐art detection methods, it exhibits superior capabilities.

In this early-stage study, we empirically characterize the runtime behavior of meaningful WebAssembly (Wasm) workloads with respect to memory allocation. We consider a variety of benchmarks and allocators, written in C and compiled to standalone Wasm, to give a broad spectrum of behavior.

The ionospheric mapping function (MF) is used to achieve mutual conversion between the vertical total electron content (VTEC) and the slant total electron content (STEC) and is vital to the application of ionospheric products. Currently, the typically used MFs consider only the effect of the ionospheric thin-layer height and signal elevation angle, whereas the effects of ionospheric spatiotemporal changes and the azimuth angle, which severely restrict the accuracy of the MF, are not considered. In this study, an ionospheric MF model with the modified Julian day (MJD), local time (LT), elevation angle, and azimuth angle of the ionospheric pierce point (IPP) as inputs is proposed. The MF model, which is named long short-term memory (LSTM)-MF, is constructed using data from the United States provided by the Massachusetts Institute of Technology (MIT)/Haystack Observatory from September 1, 2021 to April 30, 2022, and ionospheric grid products named Model VTEC are established based on the LSTM-MF model. On the test set, the root-mean-square errors (RMSEs) of the STEC projected using the LSTM-MF model in the elevation-angle ranges of 20°–40°, 40°–70°, and 70°–90° are 48.66%, 33.79%, and 11.36% higher than that of the single-layer MF (SLMF) model, respectively. On December 5, 2021, the STEC obtained by projecting the Model VTEC product using the LSTM-MF model is 43.8% higher in accuracy than that obtained by projecting MIT VTEC products using the SLMF model at the low elevation angles of 20°–50°. The LSTM-MF model proposed herein and the established VTEC product improved the STEC accuracy obtained from low-elevation-angle conversion.

Water vapor is an essential element in the hydrologic and energy cycles, as well as in the climate and atmospheric circulation on Earth. Though various techniques have been developed, water vapor is still difficult to retrieve with both high accuracy and resolution. To calibrate the biases and restrain large errors in the moderate resolution imaging spectroradiometer (MODIS) near-infrared (NIR) precipitable water vapor (PWV) in Western Europe, we propose a hybrid model that combines the spherical cap harmonic analysis (SCHA) model and the Extreme Gradient Boosting (XGBoost) model. This model includes two main steps 1) initial calibration of MODIS PWV using an SCHA model and 2) advanced calibration of the interim PWV using an XGBoost model. The results show that the hybrid model achieves an average bias of 0.0 mm, STD of 2.0 mm, and rms of 2.0 mm, increasing the MODIS PWV accuracy by 55.6% in terms of rms in Western Europe in 2020. We further demonstrate that the hybrid model outperforms the SCHA model and the XGBoost model in calibrating biases and restraining large errors. We find that the MODIS PWV typically exceeds the GNSS PWV by 0.7 mm on annual average and their difference fluctuates seasonally, varying from −4.0 mm in winter to 4.0 mm in summer. This study provides a powerful method to optimize the MODIS PWV and obtain high-quality PWV products for meteorological research and Earth observation systems.

WebAssembly, known as Wasm, is an interpreted portable bytecode execution format that is growing in popularity. In this work, we perform a simple pair of characterizations of Wasm. Statically, we compare the instruction set to other portable bytecodes like JVM and PCODE, showing that Wasm operates at a lower abstraction level than JVM, similar to PCODE. Dynamically, we study the Wasm instruction mix for a set of common benchmark applications. This investigation reveals that, like JVM, data movement operations occur most frequently in instruction traces. We conclude by discussing possible future directions for optimizing Wasm execution.

As a wireless basic unit, flexible antennas hold a wide range of applications in wearable electronics, soft robotics, and Internet of Things (IoT). However, most of the current flexible antennas are encapsulated by silicone elastomers with poor gas permeability, which severely hinders the evaporation of skin moisture and sweat. In addition, conventional rigid metals as high-frequency conductors are limited by poor elasticity and susceptibility to oxidation for on-skin application. Here, we developed a highly permeable and stretch-resistant flexible bowtie antenna that can capture changes in tensile strain and temperature. A low-impedance flexible carbon nanotube-silver (CNT-Ag) substrate was fabricated as the conductor of the antenna. By optimizing the multibeam bowed geometry and wrapping it in porous thermoplastic polyurethane (TPU) fibers, the final five-beam antenna was obtained and was able to withstand a relatively large tensile stress of 25.2 MPa, yet achieve a high vapor transmission rate of 48.2 mg cm−2 h−1. The antenna obtained an ideal impedance match at 2.28 GHz with doughnut-like radiation and a high radiation efficiency of over 85%. Furthermore, the antenna was successfully used to capture the strain in the wrist epidermis during bending and to detect thermal changes in the beaker of hot water, respectively. Finally, demonstrations of the antenna, such as permeability, radiation to the human body, and integrality in connection with flexible circuits, were carefully developed to reveal its feasibility in the real world. We expect this work to pave the way for the future establishment of epidermally flexible antennas for soft electronics.

Estimating precipitable water vapor (PWV) with high accuracy and spatial resolution is important in many disciplines. Water vapor absorption and non-absorption channels can be observed in the near-infrared (NIR) ray of the Moderate Resolution Imaging Spectroradiometer (MODIS), which can be used to retrieve PWV. However, traditional algorithms overestimate the NIR PWV in North America. This study proposes a novel NIR retrieval algorithm based on machine learning that considers land-cover types to estimate high-accuracy PWV. To do this, nonlinear models between MODIS NIR transmittance, based on the two-and three-channel ratio, and global navigation satellite system (GNSS) PWV, recorded by the SuomiNet GNSS network, are established using a backpropagation neural network (BPNN). Verification shows that the root mean square error (RMSE)/standard deviation (STD)/bias of the two-channel ratio PWV is 1.29/1.29/0.02 mm, respectively, and the improvements of RMSE and STD are 66.32% and 37.98%, respectively. The RMSE/STD/bias values of the three-channel ratio PWV are 1.29/1.29/0.02 mm, respectively, and the improvements in RMSE and STD are 68.67% and 42.31%, respectively. In addition, the surface verification of the proposed method in six land-cover types shows that both the two-and three-channel ratio methods can yield satisfactory PWV estimates. Compared with the MODIS PWV products, the proposed method yields remarkable progress.