Search

Face portrait line drawing is a unique style of art which is highly abstract and expressive. However, due to its high semantic constraints, many existing methods learn to generate portrait drawings using paired training data, which is costly and time-consuming to obtain. In this paper, we propose a novel method to automatically transform face photos to portrait drawings using unpaired training data with two new features; i.e., our method can (1) learn to generate high quality portrait drawings in multiple styles using a single network and (2) generate portrait drawings in a "new style" unseen in the training data. To achieve these benefits, we (1) propose a novel quality metric for portrait drawings which is learned from human perception, and (2) introduce a quality loss to guide the network toward generating better looking portrait drawings. We observe that existing unpaired translation methods such as CycleGAN tend to embed invisible reconstruction information indiscriminately in the whole drawings due to significant information imbalance between the photo and portrait drawing domains, which leads to important facial features missing. To address this problem, we propose a novel asymmetric cycle mapping that enforces the reconstruction information to be visible and only embedded in the selected facial regions. Along with localized discriminators for important facial regions, our method well preserves all important facial features in the generated drawings. Generator dissection further explains that our model learns to incorporate face semantic information during drawing generation. Extensive experiments including a user study show that our model outperforms state-of-the-art methods., Accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence, https://doi.org/10.1109/TPAMI.2022.3147570, code: https://github.com/yiranran/QMUPD

Recent facial image synthesis methods have been mainly based on conditional generative models. Sketch-based conditions can effectively describe the geometry of faces, including the contours of facial components, hair structures, as well as salient edges (e.g., wrinkles) on face surfaces but lack effective control of appearance, which is influenced by color, material, lighting condition, etc. To have more control of generated results, one possible approach is to apply existing disentangling works to disentangle face images into geometry and appearance representations. However, existing disentangling methods are not optimized for human face editing, and cannot achieve fine control of facial details such as wrinkles. To address this issue, we propose DeepFaceEditing, a structured disentanglement framework specifically designed for face images to support face generation and editing with disentangled control of geometry and appearance. We adopt a local-to-global approach to incorporate the face domain knowledge: local component images are decomposed into geometry and appearance representations, which are fused consistently using a global fusion module to improve generation quality. We exploit sketches to assist in extracting a better geometry representation, which also supports intuitive geometry editing via sketching. The resulting method can either extract the geometry and appearance representations from face images, or directly extract the geometry representation from face sketches. Such representations allow users to easily edit and synthesize face images, with decoupled control of their geometry and appearance. Both qualitative and quantitative evaluations show the superior detail and appearance control abilities of our method compared to state-of-the-art methods.

This paper focuses on indoor semantic segmentation based on RGB-D images. Semantic segmentation is a pixel-level classification task that has made steady progress based on fully convolutional networks (FCNs). However, we find there is still room for improvements in the following three aspects. The first is related to multi-scale feature extraction. Recent state-of-the-art works forcibly concatenate multi-scale feature representations extracted by spatial pyramid pooling, dilated convolution or other architectures, regardless of the spatial extent for each pixel. The second is regarding RGB-D modal fusion. Most successful methods treat RGB and depth as two separate modalities and force them to be joined together regardless of their different contributions to the final prediction. The final aspect is about the modeling ability of extracted features. Due to the “local grid” defined by the receptive field, the learned feature representation lacks the ability to model spatial dependencies. In addition to these modules, we design a depth estimation module to encourage the RGB network to extract more effective features. To solve the above challenges, we propose four modules to address them: scale-aware module, modality-aware module, attention module and depth estimation module. Extensive experiments on the NYU-Depth v2 and SUN RGB-D datasets demonstrate that our method is effective against RGB-D indoor semantic segmentation.

Recently, effort has been made to apply deep learning to the detection of mesh saliency. However, one major barrier is to collect a large amount of vertex-level annotation as saliency ground truth for training the neural networks. Quite a few pilot studies showed that this task is quite difficult. In this work, we solve this problem by developing a novel network trained in a weakly supervised manner. The training is end-to-end and does not require any saliency ground truth but only the class membership of meshes. Our Classification-for-Saliency CNN (CfS-CNN) employs a multi-view setup and contains a newly designed two-channel structure which integrates view-based features of both classification and saliency. It essentially transfers knowledge from 3D object classification to mesh saliency. Our approach significantly outperforms the existing state-of-the-art methods according to extensive experimental results. Also, the CfS-CNN can be directly used for scene saliency. We showcase two novel applications based on scene saliency to demonstrate its utility.

Humans have the ability to perceivekinetic depth effects, i.e., to perceived 3D shapes from 2D projections of rotating 3D objects. This process is based on a variety of visual cues such as lighting and shading effects. However, when such cues are weak or missing, perception can become faulty, as demonstrated by the famous silhouette illusion example of thespinning dancer. Inspired by this, we establish objective and subjective evaluation models of rotated 3D objects by taking their projected 2D images as input. We investigate five different cues: ambient luminance, shading, rotation speed, perspective, and color difference between the objects and background. In the objective evaluation model, we first apply 3D reconstruction algorithms to obtain an objective reconstruction quality metric, and then use quadratic stepwise regression analysis to determine weights of depth cues to represent the reconstruction quality. In the subjective evaluation model, we use a comprehensive user study to reveal correlations with reaction time and accuracy, rotation speed, and perspective. The two evaluation models are generally consistent, and potentially of benefit to inter-disciplinary research into visual perception and 3D reconstruction.

Class imbalance is a challenging problem in many classification tasks. It induces biased classification results for minority classes that contain less training samples than others. Most existing approaches aim to remedy the imbalanced number of instances among categories by resampling the majority and minority classes accordingly. However, the imbalanced level of difficulty of recognizing different categories is also crucial, especially for distinguishing samples with many classes. For example, in the task of clinical skin disease recognition, several rare diseases have a small number of training samples, but they are easy to diagnose because of their distinct visual properties. On the other hand, some common skin diseases, e.g., eczema, are hard to recognize due to the lack of special symptoms. To address this problem, we propose a self-paced balance learning (SPBL) algorithm in this paper. Specifically, we introduce a comprehensive metric termed the complexity of image category that is a combination of both sample number and recognition difficulty. First, the complexity is initialized using the model of the first pace, where the pace indicates one iteration in the self-paced learning paradigm. We then assign each class a penalty weight that is larger for more complex categories and smaller for easier ones, after which the curriculum is reconstructed by rearranging the training samples. Consequently, the model can iteratively learn discriminative representations via balancing the complexity in each pace. Experimental results on the SD-198 and SD-260 benchmark data sets demonstrate that the proposed SPBL algorithm performs favorably against the state-of-the-art methods. We also demonstrate the effectiveness of the SPBL algorithm’s generalization capacity on various tasks, such as indoor scene image recognition and object classification.

Automatic assessment of sentiment from visual content has gained considerable attention with the increasing tendency of expressing opinions online. In this paper, we solve the problem of visual sentiment analysis, which is challenging due to the high-level abstraction in the recognition process. Existing methods based on convolutional neural networks learn sentiment representations from the holistic image, despite the fact that different image regions can have different influence on the evoked sentiment. In this paper, we introduce a weakly supervised coupled convolutional network (WSCNet). Our method is dedicated to automatically selecting relevant soft proposals from weak annotations (e.g., global image labels), thereby significantly reducing the annotation burden, and encompasses the following contributions. First, WSCNet detects a sentiment-specific soft map by training a fully convolutional network with the cross spatial pooling strategy in the detection branch. Second, both the holistic and localized information are utilized by coupling the sentiment map with deep features for robust representation in the classification branch. We integrate the sentiment detection and classification branches into a unified deep framework, and optimize the network in an end-to-end way. Through this joint learning strategy, weakly supervised sentiment classification and detection benefit each other. Extensive experiments demonstrate that the proposed WSCNet outperforms the state-of-the-art results on seven benchmark datasets.

Semantic segmentation has been proposed as a tool to accelerate the processing of natural history collection images. However, developing a flexible and resilient segmentation network requires an approach for adaptation which allows processing different datasets with minimal training and validation. This paper presents a cross-validation approach designed to determine whether a semantic segmentation network possesses the flexibility required for application across different collections and institutions. Consequently, the specific objectives of cross-validating the semantic segmentation network are to (a) evaluate the effectiveness of the network for segmenting image sets derived from collections different from the one in which the network was initially trained on; and (b) test the adaptability of the segmentation network for use in other types of collections. The resilience to data variations from different institutions and the portability of the network across different types of collections are required to confirm its general applicability. The proposed validation method is tested on the Natural History Museum semantic segmentation network, designed to process entomological microscope slides. The proposed semantic segmentation network is evaluated through a series of cross-validation experiments designed to test using data from two types of collections: microscope slides (from three institutions) and herbarium sheets (from seven institutions). The main contribution of this work is the method, software and ground truth sets created for this cross-validation as they can be reused in testing similar segmentation proposals in the context of digitization of natural history collections. The cross-validation of segmentation methods should be a required step in the integration of such methods into image processing workflows for natural history collections.

This paper presents the results of SHREC’21 track: Quantifying Shape Complexity. Our goal is to investigate how good the submitted shape complexity measures are (i.e. with respect to ground truth) and investigate the relationships between these complexity measures (i.e. with respect to correlations). The dataset consists of three collections: 1800 perturbed cube and sphere models classified into 4 categories, 50 shapes inspired from the fields of architecture and design classified into 2 categories, and the data from the Princeton Segmentation Benchmark, which consists of 19 natural object categories. We evaluate the performances of the methods by computing Kendall rank correlation coefficients both between the orders produced by each complexity measure and the ground truth and between the pair of orders produced by each pair of complexity measures. Our work, being a quantitative and reproducible analysis with justified ground truths, presents an improved means and methodology for the evaluation of shape complexity.

Despite significant effort and notable success of neural style transfer, it remains challenging for highly abstract styles, in particular line drawings. In this paper, we propose APDrawingGAN++, a generative adversarial network (GAN) for transforming face photos to artistic portrait drawings (APDrawings), which addresses substantial challenges including highly abstract style, different drawing techniques for different facial features, and high perceptual sensitivity to artifacts. To address these, we propose a composite GAN architecture that consists of local networks (to learn effective representations for specific facial features) and a global network (to capture the overall content). We provide a theoretical explanation for the necessity of this composite GAN structure by proving that any GAN with a single generator cannot generate artistic styles like APDrawings. We further introduce a classification-and-synthesis approach for lips and hair where different drawing styles are used by artists, which applies suitable styles for a given input. To capture the highly abstract art form inherent in APDrawings, we address two challenging operations — (1) coping with lines with small misalignments while penalizing large discrepancy and (2) generating more continuous lines — by introducing two novel loss terms: one is a novel distance transform loss with nonlinear mapping and the other is a novel line continuity loss, both of which improve the line quality. We also develop dedicated data augmentation and pre-training to further improve results. Extensive experiments, including a user study, show that our method outperforms state-of-the-art methods, both qualitatively and quantitatively.

State-of-the-art neural style transfer methods have demonstrated amazing results by training feed-forward convolutional neural networks or using an iterative optimization strategy. The image representation used in these methods, which contains two components: style representation and content representation, is typically based on high-level features extracted from pre-trained classification networks. Because the classification networks are originally designed for object recognition, the extracted features often focus on the central object and neglect other details. As a result, the style textures tend to scatter over the stylized outputs and disrupt the content structures. To address this issue, we present a novel image stylization method that involves an additional structure representation. Our structure representation, which considers two factors: i) the global structure represented by the depth map and ii) the local structure details represented by the image edges, effectively reflects the spatial distribution of all the components in an image as well as the structure of dominant objects respectively. Experimental results demonstrate that our method achieves an impressive visual effectiveness, which is particularly significant when processing images sensitive to structure distortion, e . g . images containing multiple objects potentially at different depths, or dominant objects with clear structures.

In this paper we propose a unified framework to simultaneously discover the number of clusters and group the data points into them using subspace clustering. Real data distributed in a high-dimensional space can be disentangled into a union of low-dimensional subspaces, which can benefit various applications. To explore such intrinsic structure, state-of-the-art subspace clustering approaches often optimize a self-representation problem among all samples, to construct a pairwise affinity graph for spectral clustering. However, a graph with pairwise similarities lacks robustness for segmentation, especially for samples which lie on the intersection of two subspaces. To address this problem, we design a hyper-correlation based data structure termed as the \textit{triplet relationship}, which reveals high relevance and local compactness among three samples. The triplet relationship can be derived from the self-representation matrix, and be utilized to iteratively assign the data points to clusters. Three samples in each triplet are encouraged to be highly correlated and are considered as a meta-element during clustering, which show more robustness than pairwise relationships when segmenting two densely distributed subspaces. Based on the triplet relationship, we propose a unified optimizing scheme to automatically calculate clustering assignments. Specifically, we optimize a model selection reward and a fusion reward by simultaneously maximizing the similarity of triplets from different clusters while minimizing the correlation of triplets from same cluster. The proposed algorithm also automatically reveals the number of clusters and fuses groups to avoid over-segmentation. Extensive experimental results on both synthetic and real-world datasets validate the effectiveness and robustness of the proposed method., Comment: 13 pages, 4 figures, 6 tables

Non-rigid registration of deformed 3D shapes is a challenging and fundamental task in geometric processing, which aims to non-rigidly deform a source shape into alignment with a target shape. Current state-of-the-art methods assume deformations to be near-isometric. This assumption does not reflect real-world conditions, for example in large-scale deformation, where moderate anisotropic deformations (e.g., stretches) are common. In this paper we propose two significant changes to a typical registration pipeline to address such challenging deformations. First, we introduce a method to estimate anisotropic non-isometric deformations and incorporate this into an iterative non-rigid registration pipeline. Second, we compute additional correspondences in non-isometrically deforming regions using reliable correspondences as landmarks and prune inconsistent correspondences. We compare the performance of our proposed algorithm to several state-of-the-art methods using existing benchmarks. Experimental results show that our method outperforms existing methods.

Existing methods can generate a high dynamic range (HDR) image from a single low dynamic range (LDR) image using convolutional neural networks (CNNs). However, they are too cumbersome to run on mobile devices with limited computational resources. In this work, we design a lightweight CNN, namely LiTMNet which takes a single LDR image as input and recovers the lost information in its saturated regions to reconstruct an HDR image. To avoid trading off the reconstruction quality for efficiency, LiTMNet does not only adapt a lightweight encoder for efficient feature extraction, but also contains newly designed upsampling blocks in the decoder to alleviate artifacts and further accelerate the reconstruction. The final HDR image is produced by nonlinearly blending the network prediction and the original LDR image. Qualitative and quantitative comparisons demonstrate that LiTMNet produces HDR images of high quality comparable with the current state of the art and is faster as tested on a mobile device. Please refer to the supplementary video for additional visual results.

Despite the recent upsurge of activity in image-based non-photorealistic rendering (NPR), and in particular portrait image stylisation, due to the advent of neural style transfer, the state of performance evaluation in this field is limited, especially compared to the norms in the computer vision and machine learning communities. Unfortunately, the task of evaluating image stylisation is thus far not well defined, since it involves subjective, perceptual and aesthetic aspects. To make progress towards a solution, this paper proposes a new structured, three level, benchmark dataset for the evaluation of stylised portrait images. Rigorous criteria were used for its construction, and its consistency was validated by user studies. Moreover, a new methodology has been developed for evaluating portrait stylisation algorithms, which makes use of the different benchmark levels as well as annotations provided by user studies regarding the characteristics of the faces. We perform evaluation for a wide variety of image stylisation methods (both portrait-specific and general purpose, and also both traditional NPR approaches and neural style transfer) using the new benchmark dataset., 17 pages, 15 figures

The association between alcohol outlets and violence has long been recognised, and is commonly used to inform policing and licensing policies (such as staggered closing times and zoning). Less investigated, however, is the association between violent crime and other urban points of interest, which while associated with the city centre alcohol consumption economy, are not explicitly alcohol outlets. Here, machine learning (specifically, LASSO regression) is used to model the distribution of violent crime for the central 9 km2 of ten large UK cities. Densities of 620 different Point of Interest types (sourced from Ordnance Survey) are used as predictors, with the 10 most explanatory variables being automatically selected for each city. Cross validation is used to test generalisability of each model. Results show that the inclusion of additional point of interest types produces a more accurate model, with significant increases in performance over a baseline univariate alcohol-outlet only model. Analysis of chosen variables for city-specific models shows potential candidates for new strategies on a per-city basis, with combined-model variables showing the general trend in POI/violence association across the UK. Although alcohol outlets remain the best individual predictor of violence, other points of interest should also be considered when modelling the distribution of violence in city centres. The presented method could be used to develop targeted, city-specific initiatives that go beyond alcohol outlets and also consider other locations.

3D models are commonly used in computer vision and graphics. With the wider availability of mesh data, an efficient and intrinsic deep learning approach to processing 3D meshes is in great need. Unlike images, 3D meshes have irregular connectivity, requiring careful design to capture relations in the data. To utilize the topology information while staying robust under different triangulations, we propose to encode mesh connectivity using Laplacian spectral analysis, along with mesh feature aggregation blocks (MFABs) that can split the surface domain into local pooling patches and aggregate global information amongst them. We build a mesh hierarchy from fine to coarse using Laplacian spectral clustering, which is flexible under isometric transformations. Inside the MFABs there are pooling layers to collect local information and multi-layer perceptrons to compute vertex features of increasing complexity. To obtain the relationships among different clusters, we introduce a Correlation Net to compute a correlation matrix, which can aggregate the features globally by matrix multiplication with cluster features. Our network architecture is flexible enough to be used on meshes with different numbers of vertices. We conduct several experiments including shape segmentation and classification, and our method outperforms state-of-the-art algorithms for these tasks on the ShapeNet and COSEG datasets.

This paper presents an automatic image synthesis method to transfer the style of an example image to a content image. When standard neural style transfer approaches are used, the textures and colours in different semantic regions of the style image are often applied inappropriately to the content image, ignoring its semantic layout and ruining the transfer result. In order to reduce or avoid such effects, we propose a novel method based on automatically segmenting the objects and extracting their soft semantic masks from the style and content images, in order to preserve the structure of the content image while having the style transferred. Each soft mask of the style image represents a specific part of the style image, corresponding to the soft mask of the content image with the same semantics. Both the soft masks and source images are provided as multichannel input to an augmented deep CNN framework for style transfer which incorporates a generative Markov random field model. The results on various images show that our method outperforms the most recent techniques.

An importance measure of 3D objects inspired by human perception has a range of applications since people want computers to behave like humans in many tasks. This paper revisits a well-defined measure, distinction of 3D surface mesh, which indicates how important a region of a mesh is with respect to classification. We develop a method to compute it based on a classification network and a Markov Random Field (MRF). The classification network learns view-based distinction by handling multiple views of a 3D object. Using a classification network has an advantage of avoiding the training data problem which has become a major obstacle of applying deep learning to 3D object understanding tasks. The MRF estimates the parameters of a linear model for combining the view-based distinction maps. The experiments using several publicly accessible datasets show that the distinctive regions detected by our method are not just significantly different from those detected by methods based on handcrafted features, but more consistent with human perception. We also compare it with other perceptual measures and quantitatively evaluate its performance in the context of two applications. Furthermore, due to the view-based nature of our method, we are able to easily extend mesh distinction to 3D scenes containing multiple objects.

This paper develops a novel adaptive gradient-based block compressive sensing (AGbBCS_SP) methodology for noisy image compression and reconstruction. The AGbBCS_SP approach splits an image into blocks by maximizing their sparsity, and reconstructs images by solving a convex optimization problem. In block compressive sensing, the commonly used square block shapes cannot always produce the best results. The main contribution of our paper is to provide an adaptive method for block shape selection, improving noisy image reconstruction performance. The proposed algorithm can adaptively achieve better results by using the sparsity of pixels to adaptively select block shape. Experimental results with different image sets demonstrate that our AGbBCS_SP method is able to achieve better performance, in terms of peak signal to noise ratio (PSNR) and computational cost, than several classical algorithms.

Portrait drawing is a common form of art with high abstraction and expressiveness. Due to its unique characteristics, existing methods achieve decent results only with paired training data, which is costly and time-consuming to obtain.In this paper, we address the problem of automatic transfer from face photos to portrait drawings with unpaired training data. We observe that due to the significant imbalance of information richness between photos and drawings, existing unpaired transfer methods such as CycleGAN tends to embed invisible reconstruction information indiscriminately in the whole drawings, leading to important facial features partially missing in drawings. To address this problem, we propose a novel asymmetric cycle mapping that enforces the reconstruction information to be visible (by a truncation loss) and only embedded in selective facial regions (by a relaxed forward cycle-consistency loss). Along with localized discriminators for the eyes, nose and lips, our method well preserves all important facial features in the generated portrait drawings. By introducing a style classifier and taking the style vector into account, our method can learn to generate portrait drawings in multiple styles using a single network. Extensive experiments show that our model outperforms state-of-the-art methods.

Finding the informative subspaces of high-dimensional datasets is at the core of numerous applications in computer vision, where spectral-based subspace clustering is arguably the most widely studied method due to its strong empirical performance. Such algorithms first compute an affinity matrix to construct a self-representation for each sample using other samples as a dictionary. Sparsity and connectivity of the self-representation play important roles in effective subspace clustering. However, simultaneous optimization of both factors is difficult due to their conflicting nature, and most existing methods are designed to address only one factor. In this paper, we propose a post-processing technique to optimize both sparsity and connectivity by finding good neighbors. Good neighbors induce key connections among samples within a subspace and not only have large affinity coefficients but are also strongly connected to each other. We reassign the coefficients of the good neighbors and eliminate other entries to generate a new coefficient matrix. We show that the few good neighbors can effectively recover the subspace, and the proposed post-processing step of finding good neighbors is complementary to most existing subspace clustering algorithms. Experiments on five benchmark datasets show that the proposed algorithm performs favorably against the state-of-the-art methods with negligible additional computation cost.

In this paper we have developed a family of shape measures. All the measures from the family evaluate the degree to which a shape looks like a predefined convex polygon. A quite new approach in designing object shape based measures has been applied. In most cases such measures were defined by exploiting some shape properties. Such properties are optimized (e.g., maximized or minimized) by certain shapes and based on this, the new shape measures were defined. An illustrative example might be the shape circularity measure derived by exploiting the well-known result that the circle has the largest area among all the shapes with the same perimeter. Of course, there are many more such examples (e.g., ellipticity, linearity, elongation, and squareness measures are some of them). There are different approaches as well. In the approach applied here, no desired property is needed and no optimizing shape has to be found. We start from a desired convex polygon, and develop the related shape measure. The method also allows a tuning parameter. Thus, there is a new 2-fold family of shape measures, dependent on a predefined convex polygon, and a tuning parameter, that controls the measure's behavior. The measures obtained range over the interval (0,1] and pick the maximal possible value, equal to 1, if and only if the measured shape coincides with the selected convex polygon that was used to develop the particular measure. All the measures are invariant with respect to translations, rotations, and scaling transformations. An extension of the method leads to a family of new shape convexity measures.

Single image super-resolution is known to be an ill-posed problem, which has been studied for decades. With the developments of deep convolutional neural networks, the CNN-based single image super-resolution methods have greatly improved the quality of the generated high-resolution images. However, it is difficult for image super-resolution to make full use of the relationship between pixels in low-resolution images. To address this issue, we propose a novel multi-scale residual hierarchical dense network, which tries to find the dependencies in multi-level and multi-scale features. Especially, we apply the atrous spatial pyramid pooling, which concatenates multiple atrous convolutions with different dilation rates, and design a residual hierarchical dense structure for single image super-resolution. The atrous-spatial-pyramid-pooling module is used for learning the relationship of features at multiple scales while the residual hierarchical dense structure, which consists of several hierarchical dense blocks with skip connections, aims to adaptively detect key information from multi-level features. Meanwhile, dense features from different groups are connected in a dense approach by hierarchical dense blocks, which can adequately extract local multi-level features. The extensive experiments on benchmark datasets illustrate the superiority of our proposed method compared with the state-of-the-art methods. The super-resolution results on benchmark datasets of our method can be downloaded from https://github.com/Rainyfish/MS-RHDN, and the source code will be released upon acceptance of the paper.

This paper presents a new image synthesis method for image style transfer. For some common methods, the textures and colors in the style image are sometimes applied inappropriately to the content image, which generates artifacts. In order to improve the results, we propose a novel method based on a new strategy that combines both local and global style losses. On the one hand, a style loss function based on a local approach is used to keep the style details. On the other hand, another style loss function based on global measures is used to capture the more global structural information. Results on various images show that the proposed method reduces artifacts while faithfully transferring the style image’s characteristics and preserving the structure and color of the content image.

Mesh color edit propagation aims to propagate the color from a few color strokes to the whole mesh, which is useful for mesh colorization, color enhancement and color editing, etc. Compared with image edit propagation, luminance information is not available for 3D mesh data, so the color edit propagation is more difficult on 3D meshes than images, with far less research carried out. This paper proposes a novel solution based on sparse graph regularization. Firstly, a few color strokes are interactively drawn by the user, and then the color will be propagated to the whole mesh by minimizing a sparse graph regularized nonlinear energy function. The proposed method effectively measures geometric similarity over shapes by using a set of complementary multiscale feature descriptors, and effectively controls color bleeding via a sparse $\ell _{1}$ optimization rather than quadratic minimization used in existing work. The proposed framework can be applied for the task of interactive mesh colorization, mesh color enhancement and mesh color editing. Extensive qualitative and quantitative experiments show that the proposed method outperforms the state-of-the-art methods.

For image retrieval methods based on bag of visual words, much attention has been paid to enhancing the discriminative powers of the local features. Although retrieved images are usually similar to a query in minutiae, they may be significantly different from a semantic perspective, which can be effectively distinguished by convolutional neural networks (CNN). Such images should not be considered as relevant pairs. To tackle this problem, we propose to construct a dynamic match kernel by adaptively calculating the matching thresholds between query and candidate images based on the pairwise distance among deep CNN features. In contrast to the typical static match kernel which is independent to the global appearance of retrieved images, the dynamic one leverages the semantical similarity as a constraint for determining the matches. Accordingly, we propose a semantic-constrained retrieval framework by incorporating the dynamic match kernel, which focuses on matched patches between relevant images and filters out the ones for irrelevant pairs. Furthermore, we demonstrate that the proposed kernel complements recent methods, such as hamming embedding, multiple assignment, local descriptors aggregation, and graph-based re-ranking, while it outperforms the static one under various settings on off-the-shelf evaluation metrics. We also propose to evaluate the matched patches both quantitatively and qualitatively. Extensive experiments on five benchmark data sets and large-scale distractors validate the merits of the proposed method against the state-of-the-art methods for image retrieval.

Deformation transfer is an important research problem in geometry processing\ud and computer animation.A fundamental problem for existing deformation transfer\ud methods is to build reliable correspondences. This is challenging, especially\ud when the source and target shapes differ significantly and manual labeling is\ud typically used. We propose a novel deformation transfer method that aims at\ud minimizing user effort. We adapt a biharmonic weight deformation framework\ud which is able to produce plausible deformation even with only a few key points.\ud We then develop an automatic algorithm to identify a minimum set of key points\ud on the source model that characterizes the deformation well. While minimal user\ud effort is still needed to specify corresponding points on the target model for the\ud selected key points, our approach avoids the difficult problem of choosing key\ud points. Experimental results demonstrate that our method, despite requiring\ud little user effort, produces better deformation results than alternative solutions.\ud Keywords: shape deformation; biharmonic weights; key point selection;\ud deformation transfer

Image colorization aims to produce a natural looking color image from a given gray-scale image, which remains a challenging problem. In this paper, we propose a novel example-based image colorization method exploiting a new locality consistent sparse representation. Given a single reference color image, our method automatically colorizes the target gray-scale image by sparse pursuit. For efficiency and robustness, our method operates at the superpixel level. We extract low-level intensity features, mid-level texture features, and high-level semantic features for each superpixel, which are then concatenated to form its descriptor. The collection of feature vectors for all the superpixels from the reference image composes the dictionary. We formulate colorization of target superpixels as a dictionary-based sparse reconstruction problem. Inspired by the observation that superpixels with similar spatial location and/or feature representation are likely to match spatially close regions from the reference image, we further introduce a locality promoting regularization term into the energy formulation, which substantially improves the matching consistency and subsequent colorization results. Target superpixels are colorized based on the chrominance information from the dominant reference superpixels. Finally, to further improve coherence while preserving sharpness, we develop a new edge-preserving filter for chrominance channels with the guidance from the target gray-scale image. To the best of our knowledge, this is the first work on sparse pursuit image colorization from single reference images. Experimental results demonstrate that our colorization method outperforms the state-of-the-art methods, both visually and quantitatively using a user study.

[#17491] article suite à une conférence orale: 13th EG Euroworkshop on 3D object retrieval, 3DOR 2020, Graz, Austria, september 4-5, 2020; International audience; Proteins are natural modular objects usually composed of several domains, each domain bearing a specific function that is mediated through its surface, which is accessible to vicinal molecules. This draws attention to an understudied characteristic of protein structures: surface, that is mostly unexploited by protein structure comparison methods. In the present work, we evaluated the performance of six shape comparison methods, among which three are based on machine learning, to distinguish between 588 multi-domain proteins and to recreate the evolutionary relationships at the protein and species levels of the SCOPe database. The six groups that participated in the challenge submitted a total of 15 sets of results. We observed that the performance of all the methods significantly decreases at the species level, suggesting that shape-only protein comparison is challenging for closely related proteins. Even if the dataset is limited in size (only 588 proteins are considered whereas more than 160,000 protein structures are experimentally solved), we think that this work provides useful insights into the current shape comparison methods performance, and highlights possible limitations to large-scale applications due to the computational cost. (C) 2020 The Author(s). Published by Elsevier Ltd.

Estimating correspondence between two shapes continues to be a challenging problem in geometry processing. Most current methods assume deformation to be near-isometric, however this is often not the case. For this paper, a collection of shapes of different animals has been curated, where parts of the animals (e.g., mouths, tails & ears) correspond yet are naturally non-isometric. Ground-truth correspondences were established by asking three specialists to independently label corresponding points on each of the models with respect to a previously labelled reference model. We employ an algorithmic strategy to select a single point for each correspondence that is representative of the proposed labels. A novel technique that characterises the sparsity and distribution of correspondences is employed to measure the performance of ten shape correspondence methods.

Sketching is a natural and intuitive communication tool used for expressing concepts or ideas which are difficult to communicate through text or speech alone. Sketching is therefore used for a variety of purposes, from the expression of ideas on two-dimensional (2D) physical media, to object creation, manipulation, or deformation in three-dimensional (3D) immersive environments. This variety in sketching activities brings about a range of technologies which, while having similar scope, namely that of recording and interpreting the sketch gesture to effect some interaction, adopt different interpretation approaches according to the environment in which the sketch is drawn. In fields such as product design, sketches are drawn at various stages of the design process, and therefore, designers would benefit from sketch interpretation technologies which support these differing interactions. However, research typically focuses on one aspect of sketch interpretation and modeling such that literature on available technologies is fragmented and dispersed. In this paper, we bring together the relevant literature describing technologies which can support the product design industry, namely technologies which support the interpretation of sketches drawn on 2D media, sketch-based search interactions, as well as sketch gestures drawn in 3D media. This paper, therefore, gives a holistic view of the algorithmic support that can be provided in the design process. In so doing, we highlight the research gaps and future research directions required to provide full sketch-based interaction support.

Human visual system has the strong ability to quick assess the perceptual similarity between two facial sketches. However, existing two widely-used facial sketch metrics, e.g., FSIM and SSIM fail to address this perceptual similarity in this field. Recent study in facial modeling area has verified that the inclusion of both structure and texture has a significant positive benefit for face sketch synthesis (FSS). But which statistics are more important, and are helpful for their success? In this paper, we design a perceptual metric,called Structure Co-Occurrence Texture (Scoot), which simultaneously considers the block-level spatial structure and co-occurrence texture statistics. To test the quality of metrics, we propose three novel meta-measures based on various reliable properties. Extensive experiments demonstrate that our Scoot metric exceeds the performance of prior work. Besides, we built the first large scale (152k judgments) human-perception-based sketch database that can evaluate how well a metric is consistent with human perception. Our results suggest that "spatial structure" and "co-occurrence texture" are two generally applicable perceptual features in face sketch synthesis., Comment: Code & dataset:http://mmcheng.net/scoot/, 11 pages, ICCV 2019, First one good evaluation metric for facial sketh that consistent with human judgment. arXiv admin note: text overlap with arXiv:1804.02975

A novel shape model of multi-scale topological features is proposed which considers those features relating to connected components and holes. This is achieved by considering the \textit{persistent homology} of a pair of sublevel set functions corresponding to a pair of distance functions defined on the ambient space. The model is applicable to both single and multiple component shapes and, to the authors knowledge, is the first shape model to consider multi-scale topological features of multiple component shapes. It is demonstrated, both qualitatively and quantitatively, that the proposed model models useful multi-scale topological features and outperforms a commonly used benchmark models with respect to the task of multiple component shape retrieval.

Given a reference color image and a destination grayscale image, this paper presents a novel automatic colorization algorithm that transfers color information from the reference image to the destination image. Since the reference and destination images may contain content at different or even varying scales (due to changes of distance between objects and the camera), existing texture matching-based methods can often perform poorly. We propose a novel cross-scale texture matching method to improve the robustness and quality of the colorization results. Suitable matching scales are considered locally, which are then fused using global optimization that minimizes both the matching errors and spatial change of scales. The minimization is efficiently solved using a multi-label graph-cut algorithm. Since only low-level texture features are used, texture matching-based colorization can still produce semantically incorrect results, such as meadow appearing above the sky. We consider a class of semantic violation where the statistics of up-down relationships learned from the reference image are violated and propose an effective method to identify and correct unreasonable colorization. Finally, a novel nonlocal $\ell _{1}$ optimization framework is developed to propagate high confidence micro-scribbles to regions of lower confidence to produce a fully colorized image. Qualitative and quantitative evaluations show that our method outperforms several state-of-the-art methods.

Building on the richness of recent contributions in the field, this paper presents a state-of-the-art CNN analysis method for automating the recognition of standardised building components in modern heritage buildings. At the turn of the twentieth century manufactured building components became widely advertised for specification by architects. Consequently, a form of standardisation across various typologies began to take place. During this era of rapid economic and industrialised growth, many forms of public building were erected. This paper seeks to demonstrate a method for informing the recognition of such elements using deep learning to recognise ‘families’ of elements across a range of buildings in order to retrieve and recognise their technical specifications from the contemporary trade literature. The method is illustrated through the case of Carnegie Public Libraries in the UK, which provides a unique but ubiquitous platform from which to explore the potential for the automated recognition of manufactured standard architectural components. The aim of enhancing this knowledge base is to use the degree to which these were standardised originally as a means to inform and so support their ongoing care but also that of many other contemporary buildings. Although these libraries are numerous, they are maintained at a local level and as such, their shared challenges for maintenance remain unknown to one another. Additionally, this paper presents a methodology to indirectly retrieve useful indicators and semantics, relating to emerging HBIM families, by applying deep learning to a varied range of architectural imagery.

The need to increase global accessibility to specimens while preserving the physical specimens by reducing their handling motivates digitisation. Digitisation of natural history collections has evolved from recording of specimens’ catalogue data to including digital images and 3D models of specimens. The sheer size of the collections requires developing high throughput digitisation workflows, as well as novel acquisition systems, image standardisation, curation, preservation, and publishing. For instance, herbarium sheet digitisation workflows (and fast digitisation stations) can digitise up to 6,000 specimens per day; operating digitisation stations in parallel can increase that capacity. However, other activities of digitisation workflows still rely on manual processes which throttle the speed with which images can be published. Image quality control and information extraction from images can benefit from greater automation. This presentation explores the advantages of applying semantic segmentation (Fig. 1) to improve and automate image quality management (IQM) and information extraction from images (IEFI) of physical specimens. Two experiments were designed to determine if IQM and IEFI activities can be improved by using segments instead of full images. The time for segmenting full images needs to be considered for both IQM and IEFI. A semantic segmentation method developed by the Natural History Museum (Durrant and Livermore 2018) adapted for segmenting herbarium sheet images (Dillen et al. 2019) can process 50 images in 12 minutes. The IQM experiments evaluated the application of three quality attributes to full images and to image segments: colourfulness (Fig. 2), contrast (Fig. 3) and sharpness (Fig. 4). Evaluating colourfulness is an alternative to colour quantization algorithms such as RMSE and Delta E (Hasler and Suesstrunk 2003, Palus 2006), the method produces a value indicating if the image degrades after processing. Contrast measures the difference in luminance or colour that makes an object distinguishable. Contrast is determined by the difference in colour and brightness of the object and other objects within the same field of view (Matkovic et al. 2005, Präkel 2010). Sharpness encompasses the concepts of resolution and acutance (Bahrami and Kot 2014, Präkel 2010). Sharpness influences specimen appearance and readability of information from labels and barcodes. Evaluating the criteria on 56 barcodes and 50 colour charts segments extracted from fifty images took 34 minutes (8 minutes for the barcodes and 26 minutes for colour charts). The evaluation on the corresponding full images took 100 minutes. The processing of individual segments and full images provided results equivalent to subjective manual quality management. The IEFI experiments compared the performance of four optical character recognition (OCR) programs applied to full images (Drinkwater et al. 2014) against individual segments. The four OCR programs evaluated were Tesseract 4.X, Tesseract 3.X, Abby FineReader Engine 12, and Microsoft OneNote 2013. The test was based on a set of 250 herbarium sheet images and 1,837 segments extracted from them. The results from the experiments show that there is an average OCR speed-up of 49% when using segmented images when compared to processing times for full images (Table 1). Similarly, there was an average increase of 13% in line correctness (information from lines is ordered and not fragmented (Fig. 5, Table 2 ). Additionally, the results are useful for comparing the four OCR programs, with Tesseract 3.x offering shortest processing time, while Tesseract 4.X achieving the highest scores for line accuracy (including hand written text recognition). The results suggest that IEFI could be improved by performing OCR using segments rather than whole images, leading to faster processing and more accurate outputs. The findings support the feasibility of further automation of digitisation workflows for natural history collections. In addition to increasing the accuracy and speed of IQM and IEFI activities, the explored approaches can be packaged and published, enabling automated quality management and information extraction to be offered as a service, taking advantage of cloud platforms and workflow engines.