Drop Loss for Person Attribute Recognition With Imbalanced Noisy-Labeled Samples.

Person attribute recognition (PAR) aims to simultaneously predict multiple attributes of a person. Existing deep learning-based PAR methods have achieved impressive performance. Unfortunately, these methods usually ignore the fact that different attributes have an imbalance in the number of noisy-labeled samples in the PAR training datasets, thus leading to suboptimal performance. To address the above problem of imbalanced noisy-labeled samples, we propose a novel and effective loss called drop loss for PAR. In the drop loss, the attributes are treated differently in an easy-to-hard way. In particular, the noisy-labeled candidates, which are identified according to their gradient norms, are dropped with a higher drop rate for the harder attribute. Such a manner adaptively alleviates the adverse effect of imbalanced noisy-labeled samples on model learning. To illustrate the effectiveness of the proposed loss, we train a simple ResNet-50 model based on the drop loss and term it DropNet. Experimental results on two representative PAR tasks (including facial attribute recognition and pedestrian attribute recognition) demonstrate that the proposed DropNet achieves comparable or better performance in terms of both balanced accuracy and classification accuracy over several state-of-the-art PAR methods.

Resuspension and settlement characteristics of lake sediments amended by phosphorus inactivating materials: Implications for environmental remediation.

The classical lake internal phosphorus (P) pollution control using P-inactivating materials is typically carried out by reducing the release of soluble P from sediments to overlying water; however, particulate P loading through sediment resuspension could also cause internal P pollution for algae breeding. Therefore, based on lanthanum modified bentonite clay (Phoslock®) and drinking water treatment residues (DWTR), the effect of P inactivating materials on sediment resuspension and settlement were comprehensively investigated to assess the variations in particulate P pollution from sediment. Results showed that both materials could effectively control soluble P pollution from sediment, while both had limited effect on the supplement of particulate P to overlying water. The reason may be that hydrodynamic disturbance was the key factor regulating sediment resuspension and settlement. The disturbance induced the resuspension of different sized sediments, especially <8 µm fractions, while increasing disturbing intensities promoted resuspension of relatively larger sized sediments (e.g., <63 µm). Further analysis suggested that after resuspension and settlement, the efficiencies of reducing bioavailable P by Phoslock® in relatively large sized sediment fraction (e.g., <63 µm) were substantially less than those in relatively small sized sediment (<8 µm). Although the reducing efficacies of DWTR had limited changes in different sized sediments, the remaining bioavailable P were clearly higher in smaller sized sediments with DWTR. The different performances on bioavailable P reduction mainly resulted from the distributions of materials and original P in different sized sediments. These findings indicated the potential supplement of particulate P for algal growth during resuspension and settlement of sediments amended by P inactivating materials. Overall, understanding the sediment P bioavailability and hydraulic properties at different sizes and the lake hydrodynamic conditions is essential to develop appropriate methods to control lake internal P pollution.

Efficient removal of nitrogen and phosphorus in aqueous solutions using modified water treatment residuals-sodium alginate beads.

A high-performance sorbent, modified water treatment residuals-sodium alginate beads (WTR-SA beads), was prepared through a series of salt and combined thermal roasting composite modification between water treatment residuals and sodium alginate. The properties of modified WTR-SA beads composites were characterized by SEM-EDS, FT-IR, XRD, and BET. The adsorption performance of WTR-SA beads was investigated in removing nitrogen and phosphorus from wastewater. Compared to the unmodified WTR, the removal efficiency of nitrogen and phosphorus onto the modified WTR-SA beads was increased from 22.34 and 77.13% to 95.14 and 98.31%, respectively. The adsorption capacities of nitrogen and phosphorus onto the modified WTR-SA beads were reach a maximum of 2.52 mg/g and 6.45 mg/g, respectively. The adsorption behavior can be well described using a quasi-second-order kinetic model and Langmuir isotherm model. The thermodynamic properties of nitrogen adsorption indicated that the adsorption was spontaneous and exothermic. On the contrary, the adsorption process of phosphorus is an endothermic reaction. The adsorption of nitrogen by modified WTR-SA beads is mainly carried out through ion exchange and hydroxyl complexation, and ion exchange plays a major role in it. While, the adsorption of modified WTR-SA beads on phosphorus is affected by three actions: ligand exchange, chemical precipitation, and ion exchange, which ligand exchange is the main effect. Based on these results, it can be concluded that the modified WTR-SA beads are a high efficiency adsorbent for removing nitrogen and phosphorus from domestic and industrial wastewater.

Denitrifying microbial community with the ability to bromate reduction in a rotating biofilm-electrode reactor.

In this study, the microbial community for bromate reduction in a rotating biofilm-electrode reactor (RBER) was investigated. Continuous experiment demonstrated that the bromate reduction by an auto-hydrogenotrophic microbial community was inhibited by high concentration nitrate (50mg/L). The bacterial diversity of RBER were examined through the analyse of 16S rRNA gene sequences of clone libraries. The results showed that the bromate-reducing bacteria were phylogenetically diverse at the phylum level, representing the Firmicutes, Proteobacteria, Bacteroidetes and Actinobacteria. The relative abundances of these microbial community represented 99.1% of all phylum in the biofilms when bromate served as the sole electron acceptor. Meanwhile, the Bacillus strains became the largest phylotype and represented about 37% of the total bacteria in the biofilm, indicating that the genus Bacillus played the key role in the auto-hydrogenotrophic process. Moreover, three new bacterial genera, Exiguobacterium, Arthrobacter and Chlorobium appeared with the respective relative abundance being about 7.37%, 1.81%, and 0.52%, which might be the bromate-specific reducing bacteria.

Occurrence and risk assessment of heavy metals in sediments of the Xiangjiang River, China.

Sediment samples were collected from 22 typical metal-polluted sections along the Xiangjiang River (XJR). Spatial distribution and speciation characteristics of heavy metals in sediments of XJR were determined. Furthermore, ecological risk and enrichment degree of metals were assessed by different indices. The results showed that combined metal pollution occurred in sediments of XJR, with content range of Cd, Pb, Zn, Cu, As, Mn, Cr, and Hg reaching 2.95-29.15, 30.93-235.83, 61.50-3771.11, 9.56-81.81, 3.93-46.28, 774.83-8700.72, 10.64-65.16, and 0.13-5.09 mg kg-1, respectively. Pollution levels increased in period of industrialization but decreased after thousands of pollution enterprises were banned. Sections with serious pollution and higher risk were mainly located at Hengyang and Chang-Zhu-Tan regions (Changsha, Zhuzhou, and Xiangtan) for contaminations of Cd, As, Pb, and Hg. Values of both enrichment factor and geo-accumulation index followed the order Cd > Hg > Zn > Mn > Pb > Cu > As > Cr. Bioavailable fractions followed the order Cd (66.93 %), Zn (33.80 %), Pb (30.81 %), Mn (18.38 %), Hg (17.58 %), Cu (10.20 %), As (9.81 %), and Cr (7.65 %). Considering their bioavailability, biotoxicity, or abundance, contamination of Cd was the most dominant, and pollution of other metals should not be ignored.