Rectify representation bias in vision-language models for long-tailed recognition.

Natural data typically exhibits a long-tailed distribution, presenting great challenges for recognition tasks. Due to the extreme scarcity of training instances, tail classes often show inferior performance. In this paper, we investigate the problem within the trendy visual-language (VL) framework and find that the performance bottleneck mainly arises from the recognition confusion between tail classes and their highly correlated head classes. Building upon this observation, unlike previous research primarily emphasizing class frequency in addressing long-tailed issues, we take a novel perspective by incorporating a crucial additional factor namely class correlation. Specifically, we model the representation learning procedure for each sample as two parts, i.e., a special part that learns the unique properties of its own class and a common part that learns shared characteristics among classes. By analysis, we discover that the learning process of common representation is easily biased toward head classes. Because of the bias, the network may lean towards the biased common representation as classification criteria, rather than prioritizing the crucial information encapsulated within the specific representation, ultimately leading to recognition confusion. To solve the problem, based on the VL framework, we introduce the rectification contrastive term (ReCT) to rectify the representation bias, according to semantic hints and training status. Extensive experiments on three widely-used long-tailed datasets demonstrate the effectiveness of ReCT. On iNaturalist2018, it achieves an overall accuracy of 75.4%, surpassing the baseline by 3.6 points in a ResNet-50 visual backbone.

The Antimicrobial Peptide LJ-hep2 from Lateolabrax japonicus Exerting Activities against Multiple Pathogenic Bacteria and Immune Protection In Vivo.

Hepcidin is widely present in many kinds of fish and is an important innate immune factor. A variety of HAMP2-type hepcidins have strong antimicrobial activity and immunomodulatory functions and are expected to be developed as substitutes for antibiotics. In this study, the antimicrobial activity of Hepc2 from Japanese seabass (Lateolabrax japonicus) (designated as LJ-hep2) was investigated using its recombinant precursor protein (rLJ-hep2) expressed in Pichia pastoris and a chemically synthesized mature peptide (LJ-hep2(66-86)). The results showed that both rLJ-hep2 and synthetic LJ-hep2(66-86) displayed broad antimicrobial spectrum with potent activity against gram-negative and gram-positive bacteria, and fungi. Especially, LJ-hep2(66-86) had stronger antimicrobial activity and exhibited potent activity against several clinically isolated multidrug-resistant bacteria, including Acinetobacter baumannii, Escherichia coli, Pseudomonas aeruginosa, Klebsiella pneumoniae and Enterococcus faecium. Moreover, LJ-hep2(66-86) exerted rapid bactericidal kinetic (killed tested bacteria within 2 h), induced significant morphological changes and promoted agglutination of E. coli, P. aeruginosa and Aeromonas hydrophila. The activity of LJ-hep2(66-86) against E. coli, P. aeruginosa and A. hydrophila was stable and remained active when heated for 30 min. In addition, LJ-hep2(66-86) exhibited no cytotoxicity to the mammalian cell line HEK293T and fish cell lines (EPC and ZF4). In vivo study showed that LJ-hep2(66-86) could improve the survival rate of marine medaka (Oryzias melastigma) by about 40% under the challenge of A. hydrophila, indicating its immunoprotective function. Taken together, both rLJ-hep2 and LJ-hep2(66-86) have good prospects to be used as potential antimicrobial agents in aquaculture and medicine in the future.

Anatase-Wrapped Rutile Nanorods as an Effective Electron Collector in Hybrid Photoanodes for Visible Light-Driven Oxygen Evolution.

Arrays of single crystal TiO2 rutile nanorods (RNRs) appear highly promising as electron-collecting substrates in hybrid photoanodes as the RNRs offer direct charge carriers transport pathways, contrary to the conventional electrodes prepared from TiO2 powders that suffer from the numerous charge traps at the grain boundaries. However, the specific surface area of the nanorods is highly limited by their smooth morphology, which might be detrimental in view of utilizing the RNR as a substrate for immobilizing other functional materials. In this study, we developed a novel anatase-wrapped RNR (ARNR) material fabricated by a facile seed layer-free hydrothermal method. The ARNR comprises polycrystalline anatase nanoparticles formed on the surface of RNR, resulting in a large surface area that provides more deposition sites compared to the bare nanorods. Herein, we functionalize ARNR and RNR electrodes with polymeric carbon nitride (CNx) coupled with a CoO(OH)x cocatalyst for dioxygen evolution. The anatase wrapping of the rutile nanorod scaffold is found to be crucial for effective deposition of CNx and for improved photoanode operation in visible light-driven (λ > 420 nm) oxygen evolution, yielding a significant enhancement of photocurrent (by the factor of ∼3.7 at 1.23 V vs. RHE) and faradaic efficiency of oxygen evolution (by the factor of ∼2) as compared to photoanodes without anatase interlayer. This study thus highlights the importance of careful interfacial engineering in constructing photoelectrocatalytic systems for solar energy conversion and paves the way for the use of ARNR-based electron collectors in further hybrid and composite photochemical architectures for solar fuel production.