arXiv subjects

Dongyi Wang

Publications and source records attributed to Dongyi Wang.

24 records · Page 2Linked to original sources

arXiv2024

Neural Network Architecture Search Enabled Wide-Deep Learning (NAS-WD) for Spatially Heterogenous Property Awared Chicken Woody Breast Classification and Hardness Regression

Due to intensive genetic selection for rapid growth rates and high broiler yields in recent years, the global poultry industry has faced a challenging problem in the form of woody breast (WB) conditions. This condition has caused significant economic losses as high as $200 million annually, and the root cause of WB has yet to be identified. Human palpation is the most common method of distinguishing a WB from others. However, this method is time-consuming and subjective. Hyperspectral imaging (HSI) combined with machine learning algorithms can evaluate the WB conditions of fillets in a non-invasive, objective, and high-throughput manner. In this study, 250 raw chicken breast fillet samples (normal, mild, severe) were taken, and spatially heterogeneous hardness distribution was first considered when designing HSI processing models. The study not only classified the WB levels from HSI but also built a regression model to correlate the spectral information with sample hardness data. To achieve a satisfactory classification and regression model, a neural network architecture search (NAS) enabled a wide-deep neural network model named NAS-WD, which was developed. In NAS-WD, NAS was first used to automatically optimize the network architecture and hyperparameters. The classification results show that NAS-WD can classify the three WB levels with an overall accuracy of 95%, outperforming the traditional machine learning model, and the regression correlation between the spectral data and hardness was 0.75, which performs significantly better than traditional regression models.

cs.CV↗

arXiv2024

MEMO: Dataset and Methods for Robust Multimodal Retinal Image Registration with Large or Small Vessel Density Differences

The measurement of retinal blood flow (RBF) in capillaries can provide a powerful biomarker for the early diagnosis and treatment of ocular diseases. However, no single modality can determine capillary flowrates with high precision. Combining erythrocyte-mediated angiography (EMA) with optical coherence tomography angiography (OCTA) has the potential to achieve this goal, as EMA can measure the absolute 2D RBF of retinal microvasculature and OCTA can provide the 3D structural images of capillaries. However, multimodal retinal image registration between these two modalities remains largely unexplored. To fill this gap, we establish MEMO, the first public multimodal EMA and OCTA retinal image dataset. A unique challenge in multimodal retinal image registration between these modalities is the relatively large difference in vessel density (VD). To address this challenge, we propose a segmentation-based deep-learning framework (VDD-Reg) and a new evaluation metric (MSD), which provide robust results despite differences in vessel density. VDD-Reg consists of a vessel segmentation module and a registration module. To train the vessel segmentation module, we further designed a two-stage semi-supervised learning framework (LVD-Seg) combining supervised and unsupervised losses. We demonstrate that VDD-Reg outperforms baseline methods quantitatively and qualitatively for cases of both small VD differences (using the CF-FA dataset) and large VD differences (using our MEMO dataset). Moreover, VDD-Reg requires as few as three annotated vessel segmentation masks to maintain its accuracy, demonstrating its feasibility.

eess.IV↗

arXiv2024

High-efficiency generation of vectorial holograms with metasurfaces

Holography plays a crucial role in optics applications, but it traditionally requires complex setup and bulky devices, being unfavourable for optics integration. While metasurface-based holograms are ultra-compact and easy to realize, holographic images generated are mostly restricted to scalar ones, with a few recent attempts on vectorial holograms suffering from complex meta-structures and low efficiencies. Here, we propose and experimentally demonstrate an efficient meta-platform to generate vectorial holograms with arbitrarily designed wave fronts and polarization distributions based on ultra-compact metaatoms. Combining GS algorithm and the wave-decomposition technique, we establish a generic strategy to retrieve the optical property, i.e., the distributions of reflection phase and polarization-conversion capability of the metasurface to generate a target vectorial holographic image. We next design a series of high-efficiency and deep-subwavelength metaatoms exhibiting arbitrarily designed reflection phases and polarization-conversion capabilities, and experimentally characterize their optical properties. Based on these metaatoms, we finally realize a series of meta-holograms that can generate pre-designed vectorial holographic images upon external illuminations, and experimentally characterize their working performances. Our work provides a high-efficiency and ultra-thin platform to generate vectorial holographic images, which can find many applications in onchip photonics.

physics.optics↗

arXiv2023

Realization of a Z classified chiral-symmetric higher-order topological insulator in a coupling-inversion acoustic crystal

Higher-order topological band theory has transformed the landscape of topological phases in quantum and classical systems. Here, we experimentally demonstrate a two-dimensional (2D) higher-order topological phase (HOTP), referred to as the multiple chiral topological phase (MCTP), which is protected by a multipole chiral number (MCN). Our realization differs from previous HOTPs in that it possesses a larger-than-unity MCN, which arises when the nearest-neighbor couplings (NNCs) are weaker than long-range couplings (LRCs). Our phase has an MCN of 4, protecting the existence of 4 mid-gap topological corner modes (TCMs) at each corner. The multiple TCMs demonstrated here could lead to enhanced quantum-inspired devices for sensing and computing. Our study also highlights the rich and untapped potential of LRC manipulation for future research in topological phases.

cond-mat.mes-hall↗

arXiv2023

Gender Inequalities: Women Researchers Require More Knowledge in Specific and Experimental Topics

Gender inequalities in science have long been observed globally. Studies have demonstrated it through survey data or published literature, focusing on the interests of subjects or authors; few, however, examined the manifestation of gender inequalities on researchers' knowledge status. This study analyzes the relationship between regional and gender identities, topics, and knowledge status while revealing the female labor division in science and scientific research using online Q&A from researchers. We find that gender inequalities are merged with both regional-specific characteristics and global common patterns. Women's field and topic distribution within fields are influenced by regions, yet the prevalent topics are consistent in all regions. Women are more involved in specific topics, particularly topics about experiments with weaker levels of knowledge and they are of less assistance. To promote inequality in science, the scientific community should pay more attention to reducing the knowledge gap and encourage women to work on unexplored topics and areas.

cs.CY↗

arXiv2022

A novel illumination condition varied image dataset-Food Vision Dataset (FVD) for fair and reliable consumer acceptability predictions from food

Recent advances in artificial intelligence promote a wide range of computer vision applications in many different domains. Digital cameras, acting as human eyes, can perceive fundamental object properties, such as shapes and colors, and can be further used for conducting high-level tasks, such as image classification, and object detections. Human perceptions have been widely recognized as the ground truth for training and evaluating computer vision models. However, in some cases, humans can be deceived by what they have seen. Well-functioned human vision relies on stable external lighting while unnatural illumination would influence human perception of essential characteristics of goods. To evaluate the illumination effects on human and computer perceptions, the group presents a novel dataset, the Food Vision Dataset (FVD), to create an evaluation benchmark to quantify illumination effects, and to push forward developments of illumination estimation methods for fair and reliable consumer acceptability prediction from food appearances. FVD consists of 675 images captured under 3 different power and 5 different temperature settings every alternate day for five such days.

cs.CV↗