Search

The image-recipe cross-modal retrieval task, which retrieves the relevant recipes according to food images and vice versa, is now attracting widespread attention. There are two main challenges for image-recipe cross-modal retrieval task. Firstly, a recipe’s different components (words in a sentence, sentences in an entity, and entities in a recipe) have different weight values. If a recipe’s different components own the same weight, the recipe embeddings cannot pay more attention to the important components. As a result, the important components make less contribution to the retrieval task. Secondly, the food images have obvious properties of locality and only the local food regions matter. There are still difficulties in enhancing the discriminative local region features in the food images. To address these two problems, we propose a novel framework named Dual Cross Attention Encoders for Cross-modal Food Retrieval (DCA-Food). The proposed framework consists of a hierarchical cross attention recipe encoder (HCARE) and a cross attention image encoder (CAIE). HCARE consists of three types of cross attention modules to capture the important words in a sentence, the important sentences in an entity and the important entities in a recipe, respectively. CAIE extracts global and local region features. Then, it calculates cross attention between them to enhance the discriminative local features in the food images. We conduct the ablation studies to validate our design choices. Our proposed approach outperforms the existing approaches by a large margin on the Recipe1M dataset. Specifically, we improve the R@1 performance by +2.7 and +1.9 on the 1k and 10k testing sets, respectively.

A sensitive and simple method for detecting Cu2+ in the water source was proposed by using surface-enhanced Raman scattering spectroscopy (SERS) based on the Ag@SiO2/Au core–shell composite. The Ag@SiO2 SERS tag was synthesized by a simple approach, in which Ag nanoparticles were first embedded with Raman reporter PATP and next coated with a SiO2 shell. The Ag@SiO2 nanoparticles had strong stability even in a high-concentration salty solution, and there were no changes to their properties and appearance within one month. The Ag@SiO2/Au composite was fabricated through a controllable self-assemble process. L-cysteine was decorated on the surface of a functionalized Ag@SiO2/Au composite, as the amino and carboxyl groups of it can form coordinate covalent bond with Cu2+, which shows that the Ag@SiO2/Au composite labelled with L-cysteine has excellent performance for the detection of Cu2+ in aqueous media. In this study, the SERS detection of Cu2+ was carried out using Ag@SiO2 nanoparticles, and the limit of detection (LOD) as low as 0.1 mg/L was achieved.

A suitable population structure is the foundation for a high yield of wheat. Studying the changes in yield and population structure of different wheat rows under drip irrigation conditions can provide a theoretical basis for optimizing wheat drip irrigation pattern. In a two-year field experiment, two different water- and fertilizer-demanding spring wheat varieties (XC22 and XC44) were used to study the changes of stem and tiller dynamics, dry matter accumulation, canopy photo-synthetically active radiation (PAR) interception, and canopy apparent photosynthesis rate (CAP) under one tube serving four rows of wheat drip irrigation pattern (TR4, drip lateral spacing (DLS) = 60 cm, wheat row spacing (WRS) = 15 cm) and one tube serving six rows of wheat drip irrigation pattern (TR6, DLS = 90 cm, WRS = 15 cm; TR6L, DLS = 90 cm, WRS = 10 cm and TR6S, DLS = 80 cm, WRS = 10 cm). The results showed that under the condition of equal row spacing of 15 cm, after increasing the number of wheat rows serving by one drip irrigation tube from four (TR4, control) to six (TR6), the yields (water use efficiency) of XC22 and XC44 were lower by 11.19% and 8.63%, respectively. The reduction of yield was related to uneven population growth, specifically as follows: compared with the first wheat row (R1), at flowering stage the leaf area index (LAI) and PAR interception in the third wheat row (R3) of XC22 and XC44 were significantly decreased by 30.02%, 18.69%, 9.59%, and 14.74%, respectively. At the maturity stage, the biomass, plant height, and panicles number of tiller (TPN) in R3 were significantly decreased by 22.15%, 12.34%, 15.46%, 5.24%, 65.07%, and 42.11%, respectively. At the jointing, flowering, and milk-ripening stage, the CAP were significantly decreased by 24.65%, 22.85%, 17.06%, 14.02%, 42.14%, and 32.27%, respectively, the decrease of XC22 were all higher than that of XC44 (except for PAR interception). After the TR6 pattern was processed to narrow the wheat row spacing from 15 cm to 10 cm under the condition of the same drip tube lateral spacing (TR6L) and under the condition of shortening drip tube lateral spacing by 10 cm (TR6S), the yields in R3 of XC22 and XC44 were significantly increased by 20.07%, 18.43%, 30.39%, and 23.80%, respectively, and the increase in yields were related to the improvement of LAI, biomass, plant height, TPN, PAR interception, and increased population photosynthesis. Among the four drip irrigation patterns, for both XC22 and XC44, the yield of TR6S was the closest to that of TR4, and the yields of them were significantly higher than that of TR6 and TR6L.

As one of the most widely cultivated cotton species in China, upland cotton has moderate fiber quality and wide applicability, but its genetic basis is relatively narrow. To expand genetic diversity and improve fiber quality, in this study an introgression population (BC5S5) containing 107 lines was constructed by using G. hirsutum acc. 4105 as the recurrent parent and G. tomentosum as the donor parent. Using the specific-locus amplified fragment sequencing (SLAF-seq) strategy, 3157 high-throughput single nucleotide polymorphism (SNP) markers were obtained. Linkage analysis showed that a total of ninety-one QTLs related to fiber quality traits were detected in three environments, and the phenotypic variance explained (PVE) rates were 4.53–20.92%. Forty-six QTL (50.55%) synergistic genes were derived from G. tomentosum. Among them, qFS-A02-1 and qSCI-A02-1 were stably detected with a PVE of 9.8–16.71% and 14.78–20.92%, respectively. Within the candidate interval, Ghir_A02G012730, Ghir_A02G012790 and Ghir_A02G012830 were found to be possibly involved in cellulose and cell wall biosynthesis, with a relatively high expression during fiber development, 20 DPA and 25 DPA, which suggested that these three genes may be involved in the regulation of fiber strength traits, but their functions need further validation to determine the regulatory mechanism. Our research lays the foundation of fiber quality related to basic genetic research and breeding in cotton.

With the promotion and popularization of machine cotton-picking, more and more attention has been paid to the selection of early-maturity varieties with compact plant architecture. The type of fruit branch is one of the most important factors affecting plant architecture and early maturity of cotton. Heredity analysis of the cotton fruit branch is beneficial to the breeding of machine-picked cotton. Phenotype analysis showed that the types of fruit branches in cotton are controlled by a single recessive gene. Using an F2 population crossed with Huaxin102 (normal branch) and 04N-11 (nulliplex branch), BSA (Bulked Segregant Analysis) resequencing analysis and GhNB gene cloning in 04N-11, and allelic testing, showed that fruit branch type was controlled by the GhNB gene, located on chromosome D07. Ghnb5, a new recessive genotype of GhNB, was found in 04N-11. Through candidate gene association analysis, SNP 20_15811516_SNV was found to be associated with plant architecture and early maturity in the Xinjiang natural population. The GhNB gene, which is related to early maturity and the plant architecture of cotton, is a branch-type gene of cotton. The 20_15811516_SNV marker, obtained from the Xinjiang natural population, was used for the assisted breeding of machine-picked cotton varieties.