Generalized zero-shot learning (GZSL) aims to recognize objects from both seen and unseen classes, when only the labeled examples from seen classes are provided. Recent feature generation methods ...learn a generative model that can synthesize the missing visual features of unseen classes to mitigate the data-imbalance problem in GZSL. However, the original visual feature space is suboptimal for GZSL classification since it lacks discriminative information. To tackle this issue, we propose to integrate the generation model with the embedding model, yielding a hybrid GZSL framework. The hybrid GZSL approach maps both the real and the synthetic samples produced by the generation model into an embedding space, where we perform the final GZSL classification. Specifically, we propose a contrastive embedding (CE) for our hybrid GZSL framework. The proposed contrastive embedding can leverage not only the class-wise supervision but also the instance-wise supervision, where the latter is usually neglected by existing GZSL researches. We evaluate our proposed hybrid GZSL framework with contrastive embedding, named CE-GZSL, on five benchmark datasets. The results show that our CEGZSL method can outperform the state-of-the-arts by a significant margin on three datasets. Our codes are available on https://github.com/Hanzy1996/CE-GZSL.
Sketch-based 3D shape retrieval is a challenging task due to the large domain discrepancy between sketches and 3D shapes. Since existing methods are trained and evaluated on the same categories, they ...cannot effectively recognize the categories that have not been used during training. In this paper, we propose a novel domain disentangled generative adversarial network (DD-GAN) for zero-shot sketch-based 3D retrieval, which can retrieve the unseen categories that are not accessed during training. Specifically, we first generate domain-invariant features and domain-specific features by disentangling the learned features of sketches and 3D shapes, where the domain-invariant features are used to align with the corresponding word embeddings. Then, we develop a generative adversarial network that combines the domain-specific features of the seen categories with the aligned domain-invariant features to synthesize samples, where the synthesized samples of the unseen categories are generated by using the corresponding word embeddings. Finally, we use the synthesized samples of the unseen categories combined with the real samples of the seen categories to train the network for retrieval, so that the unseen categories can be recognized. In order to reduce the domain shift problem, we utilized unlabeled unseen samples to enhance the discrimination ability of the discriminator. With the discriminator distinguishing the generated samples from the unlabeled unseen samples, the generator can generate more realistic unseen samples. Extensive experiments on the SHREC'13 and SHREC'14 datasets show that our method significantly improves the retrieval performance of the unseen categories.
This paper reviews the first-ever image demoireing challenge that was part of the Advances in Image Manipulation (AIM) workshop, held in conjunction with ICCV 2019. This paper describes the ...challenge, and focuses on the proposed solutions and their results. Demoireing is a difficult task of removing moire patterns from an image to reveal an underlying clean image. A new dataset, called LCDMoire was created for this challenge, and consists of 10,200 synthetically generated image pairs (moire and clean ground truth). The challenge was divided into 2 tracks. Track 1 targeted fidelity, measuring the ability of demoire methods to obtain a moire-free image compared with the ground truth, while Track 2 examined the perceptual quality of demoire methods. The tracks had 60 and 39 registered participants, respectively. A total of eight teams competed in the final testing phase. The entries span the current the state-of-the-art in the image demoireing problem.
This paper reviews the first-ever image demoireing challenge that was part of the Advances in Image Manipulation (AIM) workshop, held in conjunction with ICCV 2019. This paper describes the ...challenge, and focuses on the proposed solutions and their results. Demoireing is a difficult task of removing moire patterns from an image to reveal an underlying clean image. A new dataset, called LCDMoire was created for this challenge, and consists of 10,200 synthetically generated image pairs (moire and clean ground truth). The challenge was divided into 2 tracks. Track 1 targeted fidelity, measuring the ability of demoire methods to obtain a moire-free image compared with the ground truth, while Track 2 examined the perceptual quality of demoire methods. The tracks had 60 and 39 registered participants, respectively. A total of eight teams competed in the final testing phase. The entries span the current the state-of-the-art in the image demoireing problem.
The traditional clothing classification method mainly consists of manually extracting obvious features such as color, texture and edge of the image. These artificial feature extraction methods are ...cumbersome and feature recognition rate is not high. In recent years, Deep Residual Network (ResNet) has been widely used in various fields by increasing network depth to obtain higher recognition accuracy. In this paper, the ResNet model is applied to the classification of clothing images, and on this basis, its data pooling layer is improved so that it can learn more rich features of image data. Clothing images are easy to be deformed and occluded. In this paper, a random erasing data enhancement algorithm is used to integrate and improve the model to improve the generalization ability of the ResNet model to such data. The final experimental results show that the classification accuracy of the improved residual model on clothing data in this paper is improved by 2.43%. At the same time, after integrating the random erasure data enhancement algorithm, the generalization ability of the model has been further improved.
Thermal transport is of grave importance in many high-value applications. Heat dissipation can be improved by utilizing liquid metals as thermal interface materials. Yet, liquid metals exhibit ...corrosivity towards many metals used for heat sinks, such as aluminum, and other electrical devices (i.e., copper). The compatibility of the liquid metal with the heat sink or device material as well as its long-term stability are important performance variables for thermal management systems. Herein, the compatibility of the liquid metal Galinstan, a eutectic alloy of gallium, indium, and tin, with diamond coatings and the stability of the liquid metal in this environment are scrutinized. The liquid metal did not penetrate the diamond coating nor corrode it. However, the liquid metal solidified with the progression of time, starting from the second year. After 4 years of aging, the liquid metal on all samples solidified, which cannot be explained by the dissolution of aluminum from the titanium alloy. In contrast, the solidification arose from oxidation by oxygen, followed by hydrolysis to GaOOH due to the humidity in the air. The hydrolysis led to dealloying, where In and Sn remained an alloy while Ga separated as GaOOH. This hydrolysis has implications for many devices based on gallium alloys and should be considered during the design phase of liquid metal-enabled products.
Infectious bursal disease (IBD) is a highly epidemic and immunosuppressive disease of 3- to 6-week-old chicks caused by infectious bursal disease virus (IBDV). Since 2017, there has been a notable ...increase in the isolation rates of novel variant IBDV strains in China, of which characteristic amino acid residues were different from those of early antigen variants. In this study, one IBDV strain was isolated from a farm with suspected IBD outbreak in Shandong Province, China, which was designated LY21/2. The strain LY21/2 could replicate in MC38 cells with previous culture adaption in SPF chick embryos. Phylogenetic analysis revealed that LY21/2 formed one branch with novel variant IBDVs and shared 96.8-98.6% nucleotide sequence identity with them. Moreover, LY21/2 serving as the major parent underwent the recombination event of a variant strain (19D69), while the minor parent was a very virulent strain (Harbin-1). SPF chicks inoculated with LY21/2 showed no gross clinic symptom, whereas bursal atrophy was exhibited and apoptosis was occurred in 55.21% of bursal cells. The results of histopathology and immunohistochemical staining showed that lymphocyte depletion and connective tissue hyperplasia and IBDV antigen-positive cells were observed in the bursa of LY21/2-infected chicks. Besides, DNA fragmentation was detected in the LY21/2-infected bursal tissue section by TUNEL assay. Collectivtely, these data presented analysis and evaluation of the genetic characteristics and pathogenicity of a novel variant IBDV strain. This study may help in the development of biosafety strategies for the prevention and control of IBDV in poultry.
Diabetes is considered an independent risk factor for hip fracture. In the present study, we evaluated whether perioperative glucose variability (GV) was a significant predictor of the outcomes of ...patients with diabetes after hip fracture.We analyzed the characteristics and outcomes of all patients with hip fractures admitted to our hospital between September 2008 and December 2012. Patients with diabetes were grouped into tertiles for GV, and multivariate survival analysis included age, sex, fracture type, mean fasting plasma glucose, and GV.Among the 1099 patients included in this study, 239 (21.7%) had diabetes. Patients with diabetes were more likely to develop infectious complications (5.4% vs 2.8%, P = .045), and experience mortality postoperatively (1 month: 5.5% vs 2.7%, P = .052; 12 months: 15.1% vs 8.7%, P = .006). The postoperative mortality rate was increased across the GV tertiles, and GV was an independent predictor of 1- and 12-month mortality after surgery.Patients with diabetes had poor prognoses after hip fracture. Perioperative GV is an independent predictor of mortality in patients with diabetes. Therefore, GV might be considered a valid additional parameter to consider in the management of these patients.
In 2018, a disease characterized by splenic hemorrhage and necrosis killed ducks in a duck farm in Guangxi province, China. A duck reovirus strain was isolated from the tissues of the dead ducks by ...inoculating duck embryos and BHK-21 cells. Electron microscopy of the cultured the isolate showed that the viral particles were nearly round in shape and approximately 70 nm in diameter, and they were designated DRV-GL18. Sequence analysis showed that the GL18 strain viral genome was 23,419 nt in length and had 10 dsRNA segments. Phylogenetic analysis of cDNA amplicons of segments encoding the protein σC which are outer capsid proteins showed that the isolate belongs to the branch of the epidemic strains of duck reovirus. The Recombination Detection Program (RDP) and SimPlot program analyses suggested potential genetic recombination events in the M2 segments. Pathogenicity experiments revealed that GL18 produced severe hemorrhaging in livers and necrosis in the spleen of infected SPF ducklings. A death rate of 50% in the experimental ducklings was calculated during the first 7 d, and the rest of the ducklings were observed to undergo spleen necrosis. These data suggested that GL18 is a duck reovirus isolate with severer pathogenicity, and it could be a candidate for development of vaccine. This is the first reported isolation of duck reovirus from mature ducks.
Constructing heterojunction is an attractive strategy for promoting photoelectrochemical (PEC) performance in water splitting and organic pollutant degradation. Herein, a novel porous ...BiVO4/Boron-doped Diamond (BiVO4/BDD) heterojunction photoanode containing masses of ultra-micro electrodes was successfully fabricated with an n-type BiVO4 film coated on a p-type BDD substrate by magnetron sputtering (MS). The surface structures of BiVO4 could be adjusted by changing the duration of deposition (Td). The morphologies, phase structures, electronic structures, and chemical compositions of the photoanodes were systematically characterized and analyzed. The best PEC activity with the highest current density of 1.8 mA/cm2 at 1.23 VRHE was achieved when Td was 30 min, and the sample showed the highest degradation efficiency towards tetracycline hydrochloride degradation (TCH) as well. The enhanced PEC performance was ascribed to the excellent charge transport efficiency as well as a lower carrier recombination rate, which benefited from the formation of BiVO4/BDD ultra-micro p-n heterojunction photoelectrodes and the porous structures of BiVO4. These novel photoanodes were expected to be employed in the practical PEC applications of energy regeneration and environmental management in the future.