Face Forgery Detection by 3D Decomposition and Composition Search.

Detecting digital face manipulation has attracted extensive attention due to fake media's potential risks to the public. However, recent advances have been able to reduce the forgery signals to a low magnitude. Decomposition, which reversibly decomposes an image into several constituent elements, is a promising way to highlight the hidden forgery details. In this paper, we investigate a novel 3D decomposition based method that considers a face image as the production of the interaction between 3D geometry and lighting environment. Specifically, we disentangle a face image into four graphics components including 3D shape, lighting, common texture, and identity texture, which are respectively constrained by 3D morphable model, harmonic reflectance illumination, and PCA texture model. Meanwhile, we build a fine-grained morphing network to predict 3D shapes with pixel-level accuracy to reduce the noise in the decomposed elements. Moreover, we propose a composition search strategy that enables an automatic construction of an architecture to mine forgery clues from forgery-relevant components. Extensive experiments validate that the decomposed components highlight forgery artifacts, and the searched architecture extracts discriminative forgery features. Thus, our method achieves the state-of-the-art performance.

Changes in soil prokaryotic communities and nitrogen cycling functions along a groundwater table drawdown gradient in desert wetlands.

Desert wetlands are evolving into deserts by groundwater table (GWT) drawdown. However, the changes in microbial communities and functions during the GWT drawdown are unclear, which hinders the predictive power of biogeochemical processes across the desertification. Here, 16S rRNA gene sequencing, PICRUSt2 and qPCR were used to investigate soil prokaryotic diversity, composition and nitrogen cycling gene abundance at four vegetation types [flooded swamp (FS), drained swamp (DS), desert grassland (DG), and bare sandy land (BS)] along a GWT decline gradient in the Mu Us Desert, northern China. Results showed that prokaryotic Shannon and Chao1 indexes were significantly reduced at BS than those at FS (p < 0.05). Whereas no significant difference was observed between FS, DS and DG (p > 0.05). Distinct shifts in community composition were found along the GWT decline gradient. The dominant taxa gradually changed from obligate anaerobes and eutrophic microbes to facultative anaerobes, and finally to aerobic, oligotrophic and drought-tolerant microbes. Soil moisture was the most important factor in regulating the communities. In addition, GWT drawdown inhibited the relative abundance of genes involved in nitrogen fixation, assimilatory nitrite reduction, and nitrate oxidation, but enhanced the relative abundance of genes related to denitrification, assimilated nitrate reduction, ammonia oxidation and ammonification. Thus, GWT drawdown inhibits nitrogen input potential and exacerbates nitrogen loss potential. These results help in understanding the succession characteristics of desert wetland desertification.

Changes in microbial metabolic C- and N- limitations in the rhizosphere and bulk soils along afforestation chronosequence in desertified ecosystems.

The resource acquisition strategy of soil microorganisms can be reflected by soil extracellular enzyme activity (EEA). However, there are few reports on the application of extracellular enzyme stoichiometry (EES) method to study the difference in microbial metabolic nutrient limitation between rhizosphere and bulk soil. Here, we choose the rhizosphere and bulk soils of Pinus sylvestris var. mongolica (P. sylvestris) plantations with five stand ages in the Mu Us sandy land, and analyzed the variation and differences of microbial metabolic limitation between rhizosphere and bulk soils with stand age by EES method. The results showed that the microbial metabolic C-limitation in the rhizosphere and bulk soil gradually increased with stand age. Almost all the vector angles were less than 45°, which indicated that the soil microbial metabolism was relatively limited by N rather than P. Furthermore, the microbial C- and N-limitation in rhizosphere soils were generally stronger than bulk soils. Soil physical properties (59.73%) explained most of the variations in soil EES based on the variation-partitioning analysis, followed by total nutrients (43.00%). The partial least squares path model suggested that the main driving factor for the variation of soil microbial metabolic C-limitation in the rhizosphere and bulk soils was physical properties, while the microbial N-limitation was for total nutrients. In general, the study emphasized the application of EES methods to assess the dynamic equilibrium between soil microbial resource acquisition and nutrient availability in desert ecosystems. These insights provide guidance for formulating afforestation strategies, such as nutrient management of sandy plantations.