Fine-grained Texture Research Articles

Multiscale residual gradient attention for face anti-spoofing

In the field of face anti-spoofing (FAS), how to extract the representative features to distinguish between real and spoof faces and train the corresponding deep networks are two vital issues. In this paper, we propose a simple but effective end-to-end FAS model based on an innovative texture extractor and a depth auxiliary supervision mechanism. In the feature extraction stage, we first design the residual gradient convolutions based on the redesigned gradient operators, which are used to extract fine-grained texture features. The extraction of texture features is based on multiple scales by dividing the texture differences between living and spoofing faces into three levels reasonably. Then we construct a multiscale residual gradient attention (MRGA) to obtain representative texture features from multiple levels texture features. By combining the proposed feature extractor MRGA and existing vision transformer (ViT), the MRGA-ViT is proposed to generate related semantics and obtain final classification results. In the training stage, we also propose a local depth auxiliary supervision based on a novel adjacent depth loss, which utilizes the correlation information of adjacent pixels adequately compared with traditional depth loss. The proposed MRGA-ViT model achieves competitive performance in generalization and stability ability, e.g., the ACER(%) values of intra testing on OULU-NPU database are 1.8, 2.6, 1.6 ± 1.2 and 1.9 ± 2.7 respectively, the AUC(%) of cross type testing attains 99.45 ± 0.57, the ACER(%) values of cross dataset testing are 28.1 and 36.7 respectively. Experimental results prove that the proposed model is competitive to other state-of-the-art works on generalization and stability performance.

Material Characteristics and Manufacturing Techniques Interpretation for the Casting Molds from the Silla Urban Site of Hwangnyongsa in Gyeongju, Korea

This study investigates the material characteristics of casting molds found at the Silla urban site of Hwangnyongsa in Gyeongju to interpret manufacturing techniques of casting mold for bronze used in the Unified Silla period. The casting molds appear to be earthen molds for bronze-making, with a thick ceramic wall and are classified into inner and outer mold types. The molds show a layered structure, with a silt layer thinly applied to the earthen mold surface. This layer comes into contact with melted metal during bronze manufacturing and is located on the outside of inner molds and inside of outer molds. The paste layer exhibits fine-grained texture with a bimodal structure. Thin section analysis reveals planar voids in the paste, which were made by the carbonization of rice husks used as a temper during the firing process. The silt layer is mainly composed of very fine-grained quartz and feldspars without clay and plant tempers. It is estimated that the casting molds were made by mixing soil with tempers of minerals and rice husks, applying a silt layer to the surface of the molds, and experiencing temperatures below 1,000℃. Overall, the selection of different raw materials for each layer of the molds is connected with efficiency to ensure the permeability and thermal resistance during the casting process and easy release of the bronze object from the molds.

Geochemical characterization of scoriaceous and unmelted micrometeorites from the Sør Rondane Mountains, East Antarctica: Links to chondritic parent bodies and the effects of alteration

Micrometeorites originate from the interplanetary dust complex and continuously fall to the Earth’s surface in large amounts. About 10 to 20% of micrometeorites are not melted upon reaching the Earth’s surface, preserving the primary features and characteristics of the parent material. Consequently, unmelted micrometeorites, together with scoriaceous micrometeorites, an intermediate form between cosmic spherules and unmelted micrometeorites, are pivotal in documenting the nature and evolution of interplanetary dust as well as the modifications experienced by micrometeorites during atmospheric entry. Based on their petrographic features, here we identified and characterized 64 scoriaceous and unmelted micrometeorites with diameters varying between 90 and 410 μm from fine-grained sediment sampled in the Sør Rondane Mountains of East Antarctica. Based on their size distribution, the micrometeorites from the Sør Rondane Mountains show a clear distinction between unmelted micrometeorites (<300 μm) and cosmic spherules (>400 μm) and imply an accumulation mechanism or exposure history distinct from other collections (e.g., Transantarctic Mountains). Different exposure windows, weathering processes and environmental factors (e.g., snow cover) could affect the size and composition of preserved particles.A selection of the particles (n = 49) was further characterized for geochemical composition and high-precision oxygen isotope ratios to identify potential parent bodies and document their alteration histories. About 63% of the particles, exhibiting both coarse- and fine-grained textures, derive from carbonaceous chondritic precursors. Two particles (∼4%) display anomalously 16O-poor isotopic compositions similar to that previously observed for (giant) cosmic spherules and unmelted micrometeorites, classified as “group 4” particles. These particles are thought to originate from an unidentified chondritic parent body located in a specific region of the protoplanetary disk or may have been characterized by a distinct alteration history, with recent studies linking them to CY carbonaceous chondrites. Only a single fine-grained particle (∼2%) can be assigned to ordinary chondritic parentage with confidence. The partially hydrated fine-grained matrix suggests this particle might be consistent with a Semarkona-like parent body. Approximately 10% of the studied particles exhibit extensive evidence for secondary terrestrial weathering with formation of (hydr)oxides during residence in the Antarctic environment, preventing detailed parent body identification. Ten particles (∼20%) could not be assigned to a specific parent group due to ambiguous oxygen isotope values. Overall, the parent body statistics from this study agree with those reported for different collections of a similar size fraction. Clear associations between textural groups and parent bodies could not be established. Even though unmelted micrometeorites are generally considered pristine and often do not exhibit any obvious petrographic evidence of terrestrial weathering, the chemical and isotopic data obtained here confirm that alteration can occur at the microscale and any data on unmelted particles from Antarctic subaerial collections should be evaluated with caution.

Face morphing attack detection based on high-frequency features and progressive enhancement learning.

Face morphing attacks have become increasingly complex, and existing methods exhibit certain limitations in capturing fine-grained texture and detail changes. To overcome these limitation, in this study, a detection method based on high-frequency features and progressive enhancement learning was proposed. Specifically, in this method, first, high-frequency information are extracted from the three color channels of the image to accurately capture the details and texture changes. Next, a progressive enhancement learning framework was designed to fuse high-frequency information with RGB information. This framework includes self-enhancement and interactive-enhancement modules that progressively enhance features to capture subtle morphing traces. Experiments conducted on the standard database and compared with nine classical technologies revealed that the proposed approach achieved excellent performance.

Infrared and visible image fusion based on double fluid pyramids and multi-scale gradient residual block

Infrared and visible image fusion can capture multi-modal features and generate an informative fused image. However, most existing encoder and decoder-based fusion methods cannot fully utilize the output features of each layer in the deep learning network. In most cases, there are no elaborate modules to fully extract the detail information of multi-scale feature maps. To this end, this work proposes a novel encoder and decoder-based infrared and visible image fusion network named PG-Fusion. In feature extraction part, the double fluid pyramids are designed to integrate and preserve features of each layer of the encoder in two independent data flows. Multi-layer structure and injected attention mechanisms bridge the gaps between low-level features and high-level features and promote the interaction of multi-grained features. Specifically, a multi-scale gradient residual block is devised to preserve enough fine-grained textures and details from source images, which can facilitate the extraction of discriminative information and boost the description ability of the network. In addition, sliding window comparison strategy based intensity loss function and gradient loss function are utilized to impel our network towards a final optimal solution. Extensive qualitative and quantitative experiment results illustrate that PG-Fusion achieves better fusion performance compared with eleven typical algorithms.

Depth Estimation Based on Monocular Camera Sensors in Autonomous Vehicles: A Self-supervised Learning Approach

Estimating depth from images captured by camera sensors is crucial for the advancement of autonomous driving technologies and has gained significant attention in recent years. However, most previous methods rely on stacked pooling or stride convolution to extract high-level features, which can limit network performance and lead to information redundancy. This paper proposes an improved bidirectional feature pyramid module (BiFPN) and a channel attention module (Seblock: squeeze and excitation) to address these issues in existing methods based on monocular camera sensor. The Seblock redistributes channel feature weights to enhance useful information, while the improved BiFPN facilitates efficient fusion of multi-scale features. The proposed method is in an end-to-end solution without any additional post-processing, resulting in efficient depth estimation. Experiment results show that the proposed method is competitive with state-of-the-art algorithms and preserves fine-grained texture of scene depth.

EPformer: an efficient transformer-based approach for retail product detection in fisheye images

Retail product detection in fisheye camera capture scenes frequently suffers from excessive object occlusion and deformation, as well as difficulty in distinguishing products with small fine-grained differences, so accurately classifying and localizing products in these images presents a challenge for computer vision. We propose an efficient product detection network called EPformer by fusing a visual transformer and convolutional neural network to reliably detect retail products in fisheye images. We employ a shifted window strategy for feature information interaction across windows to more precisely detect products due to the issue of dense occlusion of products. To address the issue of excessive product deformation brought on by fisheye cameras, we develop a deformation image processing module without explicit correction and embed it into the path aggregation network structure. This enables the model to efficiently capture product geometric changes and conduct feature fusion. To address the issue of differentiating fine-grained products, we design an effective coordinate squeeze-excitation (ECSE) attention module that can capture the fine-grained texture and boundary information differences between individuals in terms of spatial and channel relationships. The inability to differentiate fine-grained products can be solved by training the ECSE module in tandem with the decoupled head. The experimental results demonstrate that EPformer is a potent product detection model with a 4.9% higher mean average precision than the state-of-the-art method (YOLOX) on the fisheye product image dataset. In addition, the EPformer model can effectively detect products in fisheye images on the Jeston Xavier NX embedded device to meet the application requirements in realistic scenarios.

MIGT: Multi-modal image inpainting guided with text

In this work, we propose MIGT, a novel framework for multi-modal image inpainting that introduces textual description as guidance. We divide MIGT into three components: Coarse-to-Fine Image Inpainting Module (CFIM), Visual-Textual modalities Fusion Module (VTFM), and Multi-modal Semantic Alignment Module (MSAM). CFIM is a Unet-based inpainting model in a coarse-to-fine manner. The coarse inpainting stage produces images of rough shape and color according to the source corrupted images and the corresponding textual descriptions. The fine inpainting stage generates the final high-quality images with fine-grained textures. VTFM aims to reasonably fuse visual-textual modalities. First, we feed visual and textual features into the proposed visual-aware textual filtering mechanism to adaptively focus on desired words related to the missing areas. Then the filtered visual-aware textual features pass through an Attentional Generative Network (AGN) to obtain fusion features that are fed into the middle layers of the coarse stage for subsequent image inpainting. MSAM takes the generated image as input, reconstructing a textual description that semantically aligns the given one to guarantee the semantic consistency between the generated image and the input textual description. Extension experiments conducted on Oxford-102 flower and CUB-200–2011 bird datasets demonstrate the effectiveness of our proposed method.

PETROGRAPHIC STUDIES OF LIMESTONE IN DUKAMAJE FORMATION AT GILBEDI, GADA, SOKOTO BASIN, NIGERIA

The study was conducted in and around Gilbedi village and its environs in the Iullemmeden Basin, Sheet 4 SW, Gada local government area, Sokoto, North-West Nigeria. The research area is located in the Sokoto Basin, which is part of the Iullemmeden Basin. The study area is bounded by latitudes 13046'05"N to 13043'22"N and longitudes 5045'00"E to 5040'45"E covering 25 Km2. The mapping exercise was carried out using a topographic base map on a scale of 1:25,000. The observed Dukamaje Formation is derived from a marginal marine depositional setting and is primarily composed of shales, limestones, and mudstone with gypsum intercalation. Shales are consolidated mudstones with a fine-grained texture which is greyish. Limestone is medium to coarse-grained. It is a pale yellow to greyish in color. Petrographic observation showed the grains to be equigranular, and having interstitial quartz cement. The quartz occurs as colorless anhedral crystal with no alteration but with wavy extinction and without cleavage. Other minerals present include muscovite. Based on Folk’s classification for limestone, the limestone in the study area is called “extramicrite” because the allochems are extraclasts. The limestone also lime mud supported which is less than 10% of grains Based on Dunham’s classification schemes for limestone, the limestone of the studied Dukamaje Formation is a “mudstone”.

A new combined penetrometer-dielectric-low frequency acoustic-electrical conductivity sensor for measuring the soil physical characteristics

Development of simple and economical sensors to measure the soil physical properties is of great importance in precision agriculture. Building on a previous study with a fusion of penetrometer, dielectric and high-frequency acoustic sensors, the present work was aimed to develop a combined penetrometer-dielectric-microphone-apparent electrical conductivity (ECa) sensor to present a cost-effective instrument for measuring the soil physical properties. The multi-sensor probe consists of a cone connected to a load cell, an integrated dielectric-ECa sensor and a microphone inside the cone to record the passive sound produced at the soil-cone interface. To evaluate the multi-sensor probe, soils with different textural compositions (clay content from 142 to 537 g kg-1) were remoulded at three levels of water content and three levels of bulk density. The probe was tested in vertical penetration tests using a hydraulic loading frame. The results of acoustic signal processing using principal component analysis showed satisfactory differentiation between coarse and fine-grained texture classes. The output of the dielectric sensor was related to soil volumetric water content with R2 = 0.81 with some texture dependency. With fusion of penetrometer, dielectric and microphone sensors, R2 increased to 0.9 with RMSE= 0.031 m3 m-3. The soil relative density was predicted from data fusion of penetrometer, dielectric and microphone sensors with a regression model with R2 = 0.86 and RMSE= 3.11%. The ECa sensor showed rational responses to the concentration of NaCl solution as well as decreased the RMSE of soil water content and relative density prediction when fused with the other sensors. The multi-sensor probe developed is an economical instrument for simultaneously measuring the soil physical properties. Field evaluations are needed to further evaluate the sensor in wider range of soil texture and soil characteristics as well as in on-the-go measurement to be applied in precision agriculture.

Sem-Aug: Improving Camera-LiDAR Feature Fusion With Semantic Augmentation for 3D Vehicle Detection

Camera-LiDAR fusion provides precise distance measurements and fine-grained textures, making it a promising option for 3D vehicle detection in autonomous driving scenarios. Previous camera-LiDAR based 3D vehicle detection approaches mainly focused on employing image-based pre-trained models to fetch semantic features. However, these methods may perform inferior to the LiDAR-based ones when lacking semantic segmentation labels in autonomous driving tasks. Motivated by this observation, we propose a novel semantic augmentation method, namely Sem-Aug, to guide high-confidence camera-LiDAR fusion feature generation and boost the performance of multimodal 3D vehicle detection. The key novelty of semantic augmentation lies in the 2D segmentation mask auto-labeling, which provides supervision for semantic segmentation sub-network to mitigate the poor generalization performance of camera-LiDAR fusion. Using semantic-augmentation-guided camera-LiDAR fusion features, Sem-Aug achieves remarkable performance on the representative autonomous driving KITTI dataset compared to both the LiDAR-based baseline and previous multimodal 3D vehicle detectors. Qualitative and quantitative experiments demonstrate that Sem-Aug provides significant improvements in challenging <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Hard</i> detection scenarios caused by occlusion and truncation.

Geotechnical Properties of Some Clay Deposits in Some Parts of Southwestern Nigeria in Relation to Its Engineering Implications on Constructions

Twenty-two clay samples from Oduna, Okada, and Etsako clay deposits in Southwestern Nigeria were subjected to Geotechnical analysis to evaluate their impact on engineering, especially on construction. The results revealed that the clays deposits were majorly fine-grained texture (&lt;60%), with Plasticity ranging from 23% to 121%, Liquid Limit (25%~205%) having plasticity of medium to very high with a specific gravity from 1.93 to 2.58. Using the American association of state highway and ransportation office (AASHTO) and Unified system classification scheme (USCS), the samples fell within A-7-6 (clayey soil) indicating a subgrade that is fair to poor and within the CL-CH category of fine-grained soil with medium to very high plasticity. Compaction having maximum dry density (MDD) to optimum moisture content (OMC) ranges from 0.94 g/cm3 ~1.68 g/cm3 to 11.9%~44.5%, Triaxial result with cohesion between 0.33~35 and shear strength from 44~120 and California bearing ratio for unsoaked bottom 7.52~40, top 4.82~39.18 and soaked bottom 2.89~30.41, top 4.21~33.53. The geotechnical properties of the clay deposits do not meet the standard requirement hence the implication in engineering might be susceptibility to construction failures.

Mingling with minimal mixing: Mafic-silicic magma interactions in the Oamikaub ring complex, Namibia

The ca. 560 Ma Oamikaub ring complex (ORC) is a member of the Goas Intrusive Suite in the Damara Belt of Namibia. The volumetrically dominant granitic rocks contain a variety of microgranular magmatic enclaves (MEs), commonly with lobate and cuspate borders against their enclosing host rocks. Some also have mafic selvages and evidence for internal melt segregation. These features indicate magma mingling and deformation during host-magma flow and emplacement, and a small degree of chemical mixing with the host magmas. The spatial diversity and mineralogical characteristics of the MEs show that mechanical hybridism occurred prior to mingling at emplacement level. Despite one group of MEs forming near-linear arrays in chemical plots, the mineralogical and isotopic data show that these do not form a mixing series. A second group represents samples of chilled mafic magmas that chemically resemble the mafic rocks that form the earliest-emplaced components of the ORC. However, their fine-grained textures and isotopic characteristics show that these were a separate, later pulse of mafic magmas. There were multiple episodes of mingling and the younger pulses of mafic ORC magmas were generally more hybridised with evolved crust. Magma mingling and limited marginal hybridisation of MEs produced impressive outcrop features. However, apart from initial, source-level hybrid formation, magma mixing had little or no role in producing compositional variations in either the MEs or their host rocks. Globally, such relationships are common in enclave suites from granitic rocks, and their study can provide insights into processes and materials in magma source regions. In the present case, enriched mantle melting, crustal melting (of mafic to silicic rocks) and fractionation of tholeiitic parental magmas may all have played their parts in forming the ME suite of the ORC.

Synthesise and Characterization of Cordierite and Wollastonite Glass-Ceramics Derived from Industrial Wastes and Natural Raw Materials.

Industrial waste is one of the primary sources that harm the environment, and this topic has occupied many scientists on how to take advantage of these wastes or dispose of them and create a clean environment. By-pass cement dust is considered one of the most dangerous industrial wastes due to its fine granular size and its volatilization in the air, which causes severe environmental damage to human and animal health, and this is the reason for choosing the current research point. In this article, eight samples of glass–ceramics were prepared using by-pass cement dust and natural raw materials known as silica sand, magnesite, and kaolin. Then melted by using an electric furnace which was adjusted at a range of temperatures from 1550 to 1600 °C for 2 to 3 h; the samples were cast and were subjected to heat treatment at 1000 °C for 2 h based on the DTA results in order to produce crystalline materials. Various techniques were used to study the synthesized glass–ceramic samples, including differential thermal analysis (DTA), X-ray diffraction (XRD), scanning electron microscope (SEM), and thermal expansion coefficient (CTE). X-ray analysis showed that the phases formed through investigated glass–ceramic samples consisted mainly of β- wollastonite, parawollastonite, diopside, anorthite, and cordierite. It was noticed that β- the wollastonite phase was formed first and then turned into parawollastonite, and also, the anorthite mineral was formed at low temperatures before the formation of the diopside mineral. SEM showed that the formed microstructure turned from a coarse grain texture to a fine-grained texture, by increasing the percentage of cordierite. It also showed that the increase in time at the endothermic temperature significantly affected the crystalline texture by giving a fine-grained crystalline texture. The linear thermal expansion measurements technique used for the studied glass–ceramic samples gives thermal expansion coefficients ranging from 6.2161 × 10−6 to 2.6181 × 10−6 C−1 (in the range of 20–700 °C), and it decreased by increasing cordierite percent.

Peridotites with back-arc basin affinity exposed at the southwestern tip of the Mariana forearc

Peridotites at water depths of 3430 to 5999 m have been discovered using the submersible Shinkai6500 (dives 6K-1397 and 6K-1398) on the southwestern slope of the 139°E Ridge (11°12′N, 139°15′E), a small ridge at the southwesternmost tip of the Mariana forearc near the junction with the Yap Trench and Parece Vela Basin. The peridotites studied consist of 17 residual harzburgites and one dunite and show various textures with respect to their depths. Peridotites with coarse-grained (> 1 mm) textures were sampled from the shallowest part (3705–4042 m) of the dive area, and peridotites with fine-grained (< 0.5 mm) textures were sampled deeper (5996 m). Olivine crystal-fabrics vary with grain size, with (010)[100] A-type patterns for the coarse-grained peridotites, {0kl}[100] D-type patterns for the fine-grained peridotites, and various indistinct patterns in samples of variable grain sizes. Fine-grained peridotites with D-type olivine crystal-fabrics could result from deformation under relatively higher flow stresses, suggesting that a ductile shear zone in the lithospheric mantle could occur in the deepest part of 139°E Ridge. Spinel Cr# range from relatively low (0.36) to moderately high (up to 0.57), and correlate with Ti contents (0.07–0.45 wt.%). The trace element patterns of clinopyroxene similarly exhibit steepening slopes from the middle to the light REEs regardless of textural variations. These mineralogical and geochemical features would result from melt-rock interactions under conditions of relatively shallow lithospheric mantle, which are much more comparable with the Parece Vela Basin peridotites than the Mariana forearc peridotites. Consequently, the Parece Vela Basin mantle is more likely exposed on the inner slope of the westernmost Mariana Trench, presumably due to the collision of the Caroline Ridge.

CTE-Net: Automatic COVID-19 Lung Infection Segmentation Boosted by Low-Level Features and Fine-Grained Textures

CTE-Net: Automatic COVID-19 Lung Infection Segmentation Boosted by Low-Level Features and Fine-Grained Textures

Revisiting Deep Hyperspectral Feature Extraction Networks via Gradient Centralized Convolution

The hyperspectral images are composed of a variety of textures across the different bands which increase the spectral similarity and make it difficult to predict the pixel-wise labels without inducing additional complexity at the feature level. To extract robust and discriminative features from the different regions of land cover, the hyperspectral research community is still seeking such type of convolutions which can efficiently deal with fine-grained texture information during the feature extraction phase, which often overlook this aspect by vanilla convolution. To overcome the above shortcoming, this article proposes a generalized gradient centralized 3D convolution ( <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">G2C-Conv3D</monospace> ) operation, which is a weighted combination between the vanilla and gradient centralized 3D convolutions ( <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">GC-Conv3D</monospace> ) to extract both the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">intensity-level</i> semantic information and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">gradient-level</i> information. This can be easily plugged into the existing HSI feature extraction networks to boost the performance of accurate prediction for land-cover types. To validate the feasibility of the proposed <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">G2C-Conv3D</monospace> , we have considered the existing CNN3D, MS3DNet, ContextNet, and SSRN feature extraction models and as well as CAE3D, VAE3D, and SAE3D autoencoder (AE) networks, respectively. All these networks are embedded with <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">G2C-Conv3D</monospace> convolution to implement both generalized gradient centralized feature extraction networks (G2C-FE) and generalized gradient centralized AE networks (G2C-AE) for fine-grained spectral–spatial feature learning. In addition, <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">G2C-Conv2D</monospace> is also considered with few networks. The extensive experiments are conducted on four most widely used hyperspectral datasets i.e., IP, KSC, UH, and UP, respectively, and compared with the nine methods. The results demonstrate that the proposed <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">G2C-Conv3D</monospace> can effectively enhance the feature learning ability of the existing networks and both the qualitative and quantitative results show the superiority and effectiveness of the proposed <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">G2C-Conv3D</monospace> . The source codes will be publicly available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/danfenghong/G2C-Conv3D-HSI</uri> .

SCIFI: 3D face reconstruction via smartphone screen lighting

3D face reconstruction based on smartphones has a wide range of applications, such as face recognition, liveness detection, face animation, etc. To realize this function, these smartphones have been deployed with specialized hardware modules (infrared dot projector or depth sensor). However, it will inevitably increase the production cost of smartphones and the operation difficulty of users. In this article, we propose a smartphone screen illumination-based face reconstructIon (SCIFI) framework, which only relies on the front camera and screen lighting. Specifically, we have investigated the calibrated planar lighting to achieve the fine-grained textures in Lambertian-based reconstruction. Further, we introduce the face landmark to align multiple photographs, which aims to adjust the position mismatch caused by hand jitter. Moreover, we propose two different methods to eliminate outlier normals based on the characteristics of human face. Extensive experiments based on different environments (dark and bright), different lighting pattern (4-zones and 9-zones), and different testing subjects have validated the effectiveness and robustness of SCIFI in reconstructing the 3D face surface with a favorable surface shape as well as micro facial texture.

Insulator defect detection: A detection method of target search and cascade recognition

Aiming at the difficulties of few real samples for defective insulators and complex background of aerial images, this paper proposes a detection method based on target search and cascade recognition. Using the SINet framework, we apply fine-grained texture enhancement to different sizes of receptive fields. Through nearest-neighbor decoding and grouping reverse attention, the more recognizable features are guided to aggregate and generate a refined location area map by performing cascading purification operations. Additionally, we integrate the classification network to complete the solution. Experimental results show that the AUC value is up to 99.82%, which demonstrates the effectiveness and superiority of the proposed method on insulator defect detection.

Hierarchical accumulation network with grid attention for image super-resolution

Deep convolutional neural networks (CNNs) have recently shown promising results in single image super-resolution (SISR) due to their powerful representation ability. However, existing CNN-based SR methods mainly focus on deeper architecture design to obtain high-level semantic information, neglecting the features of intermediate layers containing fine-grained texture information and thus limiting the capacity for producing precise high-resolution images. To tackle this issue, we propose a hierarchical accumulation network (HAN) with grid attention in this paper. Specifically, a hierarchical feature accumulation (HFA) structure is proposed to accumulate outputs of intermediate layers in a grouping manner for exploiting the features of different semantic levels. Moreover, we introduce a multi-scale grid attention module (MGAM) to refine features of the same level. The MGAM employs a pyramid sampling with self-attention mechanism to efficiently model the non-local dependencies between pixel features and produces refined representations. By this means, the universal features in connection with spatial similarity and semantic levels are produced for image SR. Experimental results on five benchmark datasets with different degradation models demonstrate the superiority of our HAN in terms of quantitative metrics and visual quality.