Local Modification Research Articles

Transitions of nitrogen optical centers induced by femtosecond laser pulses in treated natural diamond

Diamond is a unique optical material with ultra-hard lattice, which properties can be altered by optically active doping defects combined with carbon interstitials and vacancies. Direct laser writing by focused ultrashort laser pulses enables local modification of color centers in diamonds with excellent three-dimensional positioning accuracy, providing unique capabilities for quantum technologies, magnetometry, laser generation and photoluminescence (PL) micromarking. To shed light on currently unknown underlying atomistic transformations, we exposed electron-irradiated and annealed natural IaAB1-type diamond with an initially high content of various nitrogen-vacancy centers (NV, H3, N3 and H4) to millions of 0.2-ps, 525-nm laser pulses focused inside the sample. The modified areas were characterized by Fourier-transform infrared (FT-IR) and ultraviolet–visible transmission spectroscopy and confocal Raman/PL microspectroscopy with excitation wavelengths of 405 nm and 532 nm. We report on a local decrease in originally strong PL yield of NV, H3 and H4 centers with a monotonic dependence on both pulse energy and exposure time, as well as a rise of H1b and H1c lines in the FT-IR spectra accompanied by a drop in the concentration of H1a center. The results indicate that laser-induced processes make possible both creation and destruction of NV, H3 and H4 centers influenced by their initial concentration and the parameters of laser modification, which can be useful for adjusting of laser writing regimes for PL micromark inscription and other local modification of diamond for photonics applications.

A Hybrid Deep Learning Model Utilizing Cross-Structural Multi-Behavioral Comparative Recommendation for Sustainable Electric Transportation Infrastructure

Ensuring the sustainability of transportation infrastructure for electric vehicles (e-trans) is increasingly imperative in the pursuit of decarbonization goals and addressing the pressing energy shortage. By prioritizing the development and maintenance of resilient e-trans platforms through the optimization of the public charging network, electric vehicle businesses can effectively meet the needs of users, thereby contributing to efforts aimed at improving environmental quality. To achieve this goal, researching the dynamics of vehicle user behaviors plays a crucial role. In this paper, we propose cross-structure multi-behavior contrastive learning for recommendation (C-MBR), which takes into account the dynamic preferences of users, and develops model profiles from the global structure module, local structure module, cross-behavior contrastive learning module, cross-structure contrastive learning module, and model prediction and optimization. C-MBR is mainly designed to learn user preferences from the diversity of users’ behaviors in the process of interacting with the project, so as to grasp the different behavioral intentions of users. The experimental and analytical research is further conducted and validated for dealing with cold start problems. The results indicate that C-MBR has a strong ability to deal with the problem of sparse data. Compared with the ablation experiment, the model performance of C-MBR is significantly enhanced, showing that the C-MBR model can fully apply the information of a global structure and local structure in cross-structure comparative learning and multi-behavioral comparative learning to further alleviate the problem of data sparsity. As a result, the e-trans infrastructure will be significantly enhanced by addressing the issue of data-driven disruption.

Assessing chemostratigraphic and biostratigraphic correlation of Ediacaran phosphorites (Bocaina, Khesen and Doushantuo Formations): Diachronic, local signals of the Neoproterozoic phosphogenic-taphonomic event

The Neoproterozoic phosphogenic event records biotic and biogeochemical changes that seem to indicate intrinsic relationships between phosphogenesis and phosphatic fossil preservation. However, chemo-litho and biostratigraphic correlations conducted on fossiliferous, phosphatic successions of the Doushantuo Formation, South China, evidence a highly heterogeneous record, featuring diachronic depositional ages, variable taxonomic diversity and inconsistent, “noisy” δ13C records. Similar detailed chemo-litho and biostratigraphic correlations targeted in high-grade phosphorite strata have not been performed along Ediacaran units elsewhere. Considering the significance and scarcity of Ediacaran high-grade phosphorite strata, establishing a comprehensive correlation framework within and between those units is fundamental to understanding Ediacaran biogeochemical and biotic evolution. This work aims to integrate new δ13C isotope stratigraphy, and existing Doushantuo-Pertatataka acritarch assemblages and geochronological records from Bocaina (Brazil), Khesen (Mongolia) and Doushantuo (China) Formations. Phosphate rich strata in those units feature inconsistent δ13C signatures characterized by negative values showing large bed-to-bed variations, possibly reflecting local modification of original marine signals by phosphogenesis, suggesting caution against the use in global/regional chemostratigraphic correlation. The record of phosphatized acritarch assemblages shows contrastive taxonomic diversities, and a small number of common taxa including Megasphaera. Whereas the presence of biozonal Doushantuo Formation taxa on Khsesen Fm, resonates with growing views of strong environmental resilience and large temporal ranges of acanthomorph acritarchs throughout the Ediacaran-early Cambrian, the scarcity of similar taxa in the Bocaina Fm may question the reliability of Doushantuo Fm acritarch biozones in other paleogeographic or paleoenvironmental scenarios. Alternatively, the low taxonomic diversity recorded in Bocaina and Khesen Fms may also reflect the effects of the late Ediacaran decline in acritarch diversity. Further, combined evidence of selective paleoenvironmental phosphatization, local geochemical signals, narrow time spans and heterogeneous taxonomic diversity found in phosphatized DPK acanthomorph assemblages indicate the local nature of Ediacaran phosphatic fossil preservation events. Finally, evidence for diachronic preservation between high-grade phosphatic strata and phosphatized microfossils show non-causal relationships between Ediacaran phosphogenesis and phosphatic fossil preservation, highlighting the punctuated, intrinsic local nature of those events.

Dynamic localization of the chromosomal passenger complex in trypanosomes is controlled by the orphan kinesins KIN-A and KIN-B.

The chromosomal passenger complex (CPC) is an important regulator of cell division, which shows dynamic subcellular localization throughout mitosis, including kinetochores and the spindle midzone. In traditional model eukaryotes such as yeasts and humans, the CPC consists of the catalytic subunit Aurora B kinase, its activator INCENP, and the localization module proteins Borealin and Survivin. Intriguingly, Aurora B and INCENP as well as their localization pattern are conserved in kinetoplastids, an evolutionarily divergent group of eukaryotes that possess unique kinetochore proteins and lack homologs of Borealin or Survivin. It is not understood how the kinetoplastid CPC assembles nor how it is targeted to its subcellular destinations during the cell cycle. Here, we identify two orphan kinesins, KIN-A and KIN-B, as bona fide CPC proteins in Trypanosoma brucei, the kinetoplastid parasite that causes African sleeping sickness. KIN-A and KIN-B form a scaffold for the assembly of the remaining CPC subunits. We show that the C-terminal unstructured tail of KIN-A interacts with the KKT8 complex at kinetochores, while its N-terminal motor domain promotes CPC translocation to spindle microtubules. Thus, the KIN-A:KIN-B complex constitutes a unique 'two-in-one' CPC localization module, which directs the CPC to kinetochores from S phase until metaphase and to the central spindle in anaphase. Our findings highlight the evolutionary diversity of CPC proteins and raise the possibility that kinesins may have served as the original transport vehicles for Aurora kinases in early eukaryotes.

Dynamic localization of the chromosomal passenger complex in trypanosomes is controlled by the orphan kinesins KIN-A and KIN-B

The chromosomal passenger complex (CPC) is an important regulator of cell division, which shows dynamic subcellular localization throughout mitosis, including kinetochores and the spindle midzone. In traditional model eukaryotes such as yeasts and humans, the CPC consists of the catalytic subunit Aurora B kinase, its activator INCENP, and the localization module proteins Borealin and Survivin. Intriguingly, Aurora B and INCENP as well as their localization pattern are conserved in kinetoplastids, an evolutionarily divergent group of eukaryotes that possess unique kinetochore proteins and lack homologs of Borealin or Survivin. It is not understood how the kinetoplastid CPC assembles nor how it is targeted to its subcellular destinations during the cell cycle. Here, we identify two orphan kinesins, KIN-A and KIN-B, as bona fide CPC proteins in Trypanosoma brucei, the kinetoplastid parasite that causes African sleeping sickness. KIN-A and KIN-B form a scaffold for the assembly of the remaining CPC subunits. We show that the C-terminal unstructured tail of KIN-A interacts with the KKT8 complex at kinetochores, while its N-terminal motor domain promotes CPC translocation to spindle microtubules. Thus, the KIN-A:KIN-B complex constitutes a unique ‘two-in-one’ CPC localization module, which directs the CPC to kinetochores from S phase until metaphase and to the central spindle in anaphase. Our findings highlight the evolutionary diversity of CPC proteins and raise the possibility that kinesins may have served as the original transport vehicles for Aurora kinases in early eukaryotes.

A Formalism of F-modules for Rings with Complete Local Finite F-Representation Type

Abstract We develop a formalism of unit $F$-modules in the style of Lyubeznik and Emerton-Kisin for rings that have finite $F$-representation type after localization and completion at every prime ideal. As applications, we show that if $R$ is such a ring then the iterated local cohomology modules $H^{n_{1}}_{I_{1}} \circ \cdots \circ H^{n_{s}}_{I_{s}}(R)$ have finitely many associated primes, and that all local cohomology modules $H^{n}_{I}(R / gR)$ have closed support when $g$ is a nonzerodivisor on $R$.

A Comprehensive Exploration of Diverse Skin Cell Types in the Limb of the Desert Tortoise (Testudo graeca) through Light, Transmission, Scanning Electron Microscopy, and Immunofluorescence Techniques

The Greek tortoise, inhabiting harsh desert environments, provides a compelling case for investigating skin adaptations to extreme conditions. We have utilized light microscopy, scanning electron microscopy (SEM), transmission electron microscopy (TEM), and immunofluorescence analysis to describe the structure of the arid-adapted limb skin in the Greek tortoise. Our aim was to identify the cell types that reflect the skin adaptation of this tortoise to arid conditions. Utilizing seven antibodies, we localized and elucidated the functions of various skin cells, shedding light on how the tortoise adapts to adverse environmental conditions. Our findings unveiled numerous scales on the limbs, varying in size and color, acting as protective armor against abrasions, bites, and other potential threats in their rocky habitats. The epidermis comprises four layers: stratum basalis, stratum spinosum, peri-corneous layer, and stratum corneum. Cytokeratin 14 (CK14) was explicitly detected in the basal layer of the epidermis, suggesting a role in maintaining epidermal integrity and cellular function. Langerhans cells were observed between epidermal cells filled with ribosomes and Birbeck granules. Numerous dendritic-shaped Langerhans cells revealed through E-Cadherin signify strong immunity in tortoises' skin. Melanophores were identified using the Melan-A antibody, labeling the cytoplasm, and the SOX10 antibody, labeling the nucleus, providing comprehensive insights into melanophores morphology and distribution. Two types of melanophores were found: dendritic below the stratum basalis of the epidermis and clustered oval melanophores in the deep dermal layer. Varied melanophores distribution resulted in a spotted skin pattern, potentially offering adaptive camouflage and protection against environmental challenges. Numerous myofibroblasts were discerned through alpha-smooth actin (α-SMA) expression, indicating that the Greek tortoise's skin possesses a robust tissue repair and remodeling capacity. B-cell lymphocytes detected via CD20 immunostaining exhibited sporadic distribution in the dermis, concentrating in lymphoid aggregates and around vessels, implying potential roles in local immune responses and inflammation modulation. Employing Tom20 to identify skin cells with abundant mitochondria revealed a notable presence in melanophores and the basal layer of the epidermis, suggesting high metabolic activity in these cell types and potentially influencing cellular functions. These findings contribute to our comprehension of tortoise skin anatomy and physiology, offering insights into the remarkable adaptations of this species finely tuned to their specific environmental habitats.

Another role of CO-formation catalyst in acidic tandem CO2 electroreduction: Local pH modulator

Another role of CO-formation catalyst in acidic tandem CO2 electroreduction: Local pH modulator

LIFNS: Design of a novel Lidar-IMU fusion navigation system for AGVs in smart factories

Traditional Automated Guided Vehicle (AGV) robots commonly employ 2D plane navigation systems. However, with the expansion of AGV application scenarios and the increasing complexity of structures, the limitations of existing 2D navigation systems have become more pronounced, rendering them inadequate for addressing these challenges. To tackle this issue, this paper proposes a novel navigation system suitable for AGV robots composed of fused Lidar and Inertial Measurement Unit (IMU), named as Lidar-IMU Fusion Navigation System (LIFNS). LIFNS primarily comprises mapping, localization, and fused path planning modules. In the mapping module, a 3D point cloud map for precise 3D localization and a 2D grid map for path planning are constructed using multiple-line Lidar and IMU. For the localization module, a fusion localization approach is introduced, combining IMU data with Normal Distributions Transform (NDT) point cloud registration through Unscented Kalman Filtering (UKF). Finally, global path planning is executed using the A* algorithm on the grid map, while local path planning utilizes the Timed-Elastic Band (TEB) algorithm. The effectiveness and universality of LIFNS are validated through simulation experiments and real-world deployment tests. The experimental results demonstrate that LIFNS achieves centimeter-level accuracy in both mapping and localization, effectively alleviating the issues of low precision and significant limitations present in traditional AGV robots. This positions LIFNS with promising applications in enclosed settings such as smart factories, industrial parks, and healthcare facilities.

Development of an in situ alloying method for high-performance welding processes to achieve an LTT effect by local modification of the alloy content

One possible option for increasing the fatigue strength of welded joints is the use of so-called low transformation temperature (LTT) alloys. The aim is to introduce residual compressive stresses into the weld to counteract crack initiation and propagation. Until now, there has been no application of an LTT effect to high-performance welding processes such as the laser beam submerged arc hybrid welding process (LUPuS hybrid). First, the LUPuS hybrid single-wire process was further developed into the LUPuS tandem hybrid process. This makes it possible to equip the two submerged arc welding torches with different commercially available filler wires. The aim of the work is to further develop the LUPuS tandem hybrid welding process to enable the use of the LTT effect. The in situ alloying process for obtaining the LTT effect from commercially available material combinations was extended to the two-wire process. The alloy obtained was investigated by means of energy dispersive x-ray spectroscopy and hardness measurements and the influence on residual stresses was determined by the borehole method supported by electronic speckle pattern interferometry.

From Lewis Acid to Lewis Base by La3+-to-Y3+ Substitution in α-YB5O9: Local Structure Modification Induced Lewis Basicity.

Different from the common perspective of average structure, we propose that the locally elongated metal-oxygen bonds induced by La3+-to-Y3+ substitution to a Lewis acid α-YB5O9 generate medium-strength basic sites. Experimentally, NH3- and CO2-TPD experiments prove that the La3+ doping of α-Y1-xLaxB5O9 (0 ≤ x ≤ 0.24) results in the emergence of new medium-strength basic sites and the increasing La3+ concentration modifies the number, not the strength, of the acidic and basic sites. The catalytic IPA conversion exhibits a reversal of the product selectivity, i.e., from 93% of propylene for α-YB5O9 to ∼90% of acetone for α-Y0.76La0.24B5O9, which means the La3+ doping gradually turns the solid from a Lewis acid to a Lewis base. Besides, α-Y0.76RE0.24B5O9 (RE = Ce, Eu, Gd, Tm) compounds were prepared to consolidate the above conjecture, where the acetone selectivity exhibits a linear dependence on the ionic radius (or electronegativity). This work suggests that the substitution-induced local structure change deserves more attention.

Revisiting Graph-Based Fraud Detection in Sight of Heterophily and Spectrum

Graph-based fraud detection (GFD) can be regarded as a challenging semi-supervised node binary classification task. In recent years, Graph Neural Networks (GNN) have been widely applied to GFD, characterizing the anomalous possibility of a node by aggregating neighbor information. However, fraud graphs are inherently heterophilic, thus most of GNNs perform poorly due to their assumption of homophily. In addition, due to the existence of heterophily and class imbalance problem, the existing models do not fully utilize the precious node label information. To address the above issues, this paper proposes a semi-supervised GNN-based fraud detector SEC-GFD. This detector includes a hybrid filtering module and a local environmental constraint module, the two modules are utilized to solve heterophily and label utilization problem respectively. The first module starts from the perspective of the spectral domain, and solves the heterophily problem to a certain extent. Specifically, it divides the spectrum into various mixed-frequency bands based on the correlation between spectrum energy distribution and heterophily. Then in order to make full use of the node label information, a local environmental constraint module is adaptively designed. The comprehensive experimental results on four real-world fraud detection datasets denote that SEC-GFD outperforms other competitive graph-based fraud detectors. We release our code at https://github.com/Sunxkissed/SEC-GFD.

Recasting Regional Lighting for Shadow Removal

Removing shadows requires an understanding of both lighting conditions and object textures in a scene. Existing methods typically learn pixel-level color mappings between shadow and non-shadow images, in which the joint modeling of lighting and object textures is implicit and inadequate. We observe that in a shadow region, the degradation degree of object textures depends on the local illumination, while simply enhancing the local illumination cannot fully recover the attenuated textures. Based on this observation, we propose to condition the restoration of attenuated textures on the corrected local lighting in the shadow region. Specifically, We first design a shadow-aware decomposition network to estimate the illumination and reflectance layers of shadow regions explicitly. We then propose a novel bilateral correction network to recast the lighting of shadow regions in the illumination layer via a novel local lighting correction module, and to restore the textures conditioned on the corrected illumination layer via a novel illumination-guided texture restoration module. We further annotate pixel-wise shadow masks for the public SRD dataset, which originally contains only image pairs. Experiments on three benchmarks show that our method outperforms existing state-of-the-art shadow removal methods. Project page in: yuhaoliu7456.github.io/RRL-Net.

Privileged Prior Information Distillation for Image Matting

Performance of trimap-free image matting methods is limited when trying to decouple the deterministic and undetermined regions, especially in the scenes where foregrounds are semantically ambiguous, chromaless, or high transmittance. In this paper, we propose a novel framework named Privileged Prior Information Distillation for Image Matting (PPID-IM) that can effectively transfer privileged prior environment-aware information to improve the performance of trimap-free students in solving hard foregrounds. The prior information of trimap regulates only the teacher model during the training stage, while not being fed into the student network during actual inference. To achieve effective privileged cross-modality (i.e. trimap and RGB) information distillation, we introduce a Cross-Level Semantic Distillation (CLSD) module that reinforces the students with more knowledgeable semantic representations and environment-aware information. We also propose an Attention-Guided Local Distillation module that efficiently transfers privileged local attributes from the trimap-based teacher to trimap-free students for the guidance of local-region optimization. Extensive experiments demonstrate the effectiveness and superiority of our PPID on image matting. The code will be released soon.

BCLNet: Bilateral Consensus Learning for Two-View Correspondence Pruning

Correspondence pruning aims to establish reliable correspondences between two related images and recover relative camera motion. Existing approaches often employ a progressive strategy to handle the local and global contexts, with a prominent emphasis on transitioning from local to global, resulting in the neglect of interactions between different contexts. To tackle this issue, we propose a parallel context learning strategy that involves acquiring bilateral consensus for the two-view correspondence pruning task. In our approach, we design a distinctive self-attention block to capture global context and parallel process it with the established local context learning module, which enables us to simultaneously capture both local and global consensuses. By combining these local and global consensuses, we derive the required bilateral consensus. We also design a recalibration block, reducing the influence of erroneous consensus information and enhancing the robustness of the model. The culmination of our efforts is the Bilateral Consensus Learning Network (BCLNet), which efficiently estimates camera pose and identifies inliers (true correspondences). Extensive experiments results demonstrate that our network not only surpasses state-of-the-art methods on benchmark datasets but also showcases robust generalization abilities across various feature extraction techniques. Noteworthily, BCLNet obtains significant improvement gains over the second best method on unknown outdoor dataset, and obviously accelerates model training speed.

Multi-Modal Disordered Representation Learning Network for Description-Based Person Search

Description-based person search aims to retrieve images of the target identity via textual descriptions. One of the challenges for this task is to extract discriminative representation from images and descriptions. Most existing methods apply the part-based split method or external models to explore the fine-grained details of local features, which ignore the global relationship between partial information and cause network instability. To overcome these issues, we propose a Multi-modal Disordered Representation Learning Network (MDRL) for description-based person search to fully extract the visual and textual representations. Specifically, we design a Cross-modality Global Feature Learning Architecture to learn the global features from the two modalities and meet the demand of the task. Based on our global network, we introduce a Disorder Local Learning Module to explore local features by a disordered reorganization strategy from both visual and textual aspects and enhance the robustness of the whole network. Besides, we introduce a Cross-modality Interaction Module to guide the two streams to extract visual or textual representations considering the correlation between modalities. Extensive experiments are conducted on two public datasets, and the results show that our method outperforms the state-of-the-art methods on CUHK-PEDES and ICFG-PEDES datasets and achieves superior performance.

LGMRec: Local and Global Graph Learning for Multimodal Recommendation

The multimodal recommendation has gradually become the infrastructure of online media platforms, enabling them to provide personalized service to users through a joint modeling of user historical behaviors (e.g., purchases, clicks) and item various modalities (e.g., visual and textual). The majority of existing studies typically focus on utilizing modal features or modal-related graph structure to learn user local interests. Nevertheless, these approaches encounter two limitations: (1) Shared updates of user ID embeddings result in the consequential coupling between collaboration and multimodal signals; (2) Lack of exploration into robust global user interests to alleviate the sparse interaction problems faced by local interest modeling. To address these issues, we propose a novel Local and Global Graph Learning-guided Multimodal Recommender (LGMRec), which jointly models local and global user interests. Specifically, we present a local graph embedding module to independently learn collaborative-related and modality-related embeddings of users and items with local topological relations. Moreover, a global hypergraph embedding module is designed to capture global user and item embeddings by modeling insightful global dependency relations. The global embeddings acquired within the hypergraph embedding space can then be combined with two decoupled local embeddings to improve the accuracy and robustness of recommendations. Extensive experiments conducted on three benchmark datasets demonstrate the superiority of our LGMRec over various state-of-the-art recommendation baselines, showcasing its effectiveness in modeling both local and global user interests.

Removing Interference and Recovering Content Imaginatively for Visible Watermark Removal

Visible watermarks, while instrumental in protecting image copyrights, frequently distort the underlying content, complicating tasks like scene interpretation and image editing. Visible watermark removal aims to eliminate the interference of watermarks and restore the background content. However, existing methods often implement watermark component removal and background restoration tasks within a singular branch, leading to residual watermarks in the predictions and ignoring cases where watermarks heavily obscure the background. To address these limitations, this study introduces the Removing Interference and Recovering Content Imaginatively (RIRCI) framework. RIRCI embodies a two-stage approach: the initial phase centers on discerning and segregating the watermark component, while the subsequent phase focuses on background content restoration. To achieve meticulous background restoration, our proposed model employs a dual-path network capable of fully exploring the intrinsic background information beneath semi-transparent watermarks and peripheral contextual information from unaffected regions. Moreover, a Global and Local Context Interaction module is built upon multi-layer perceptrons and bidirectional feature transformation for comprehensive representation modeling in the background restoration phase. The efficacy of our approach is empirically validated across two large-scale datasets, and our findings reveal a marked enhancement over existing watermark removal techniques.

Towards Compact 3D Representations via Point Feature Enhancement Masked Autoencoders

Learning 3D representation plays a critical role in masked autoencoder (MAE) based pre-training methods for point cloud, including single-modal and cross-modal based MAE. Specifically, although cross-modal MAE methods learn strong 3D representations via the auxiliary of other modal knowledge, they often suffer from heavy computational burdens and heavily rely on massive cross-modal data pairs that are often unavailable, which hinders their applications in practice. Instead, single-modal methods with solely point clouds as input are preferred in real applications due to their simplicity and efficiency. However, such methods easily suffer from limited 3D representations with global random mask input. To learn compact 3D representations, we propose a simple yet effective Point Feature Enhancement Masked Autoencoders (Point-FEMAE), which mainly consists of a global branch and a local branch to capture latent semantic features. Specifically, to learn more compact features, a share-parameter Transformer encoder is introduced to extract point features from the global and local unmasked patches obtained by global random and local block mask strategies, followed by a specific decoder to reconstruct. Meanwhile, to further enhance features in the local branch, we propose a Local Enhancement Module with local patch convolution to perceive fine-grained local context at larger scales. Our method significantly improves the pre-training efficiency compared to cross-modal alternatives, and extensive downstream experiments underscore the state-of-the-art effectiveness, particularly outperforming our baseline (Point-MAE) by 5.16%, 5.00%, and 5.04% in three variants of ScanObjectNN, respectively. Code is available at https://github.com/zyh16143998882/AAAI24-PointFEMAE.

RetouchFormer: Semi-supervised High-Quality Face Retouching Transformer with Prior-Based Selective Self-Attention

Face retouching is to beautify a face image, while preserving the image content as much as possible. It is a promising yet challenging task to remove face imperfections and fill with normal skin. Generic image enhancement methods are hampered by the lack of imperfection localization, which often results in incomplete removal of blemishes at large scales. To address this issue, we propose a transformer-based approach, RetouchFormer, which simultaneously identify imperfections and synthesize realistic content in the corresponding regions. Specifically, we learn a latent dictionary to capture the clean face priors, and predict the imperfection regions via a reconstruction-oriented localization module. Also based on this, we can realize face retouching by explicitly suppressing imperfections in our selective self-attention computation, such that local content will be synthesized from normal skin. On the other hand, multi-scale feature tokens lead to increased flexibility in dealing with the imperfections at various scales. The design elements bring greater effectiveness and efficiency. RetouchFormer outperforms the advanced face retouching methods and synthesizes clean face images with high fidelity in our list of extensive experiments performed.