Facial beauty prediction (FBP) is a significant visual recognition problem to make assessment of facial attractiveness that consistent human perception. To tackle this problem, various data-driven models, especially state-of-the-art deep learning techniques, were introduced, and benchmark dataset become one the essential elements achieve FBP. Previous works have formulated as specific supervised classification, regression or ranking, which indicates FBP intrinsically computation with...

Video-based human action recognition is one of the most important and challenging areas research in field computer vision. Human has found many pragmatic applications video surveillance, human-computer interaction, entertainment, autonomous driving, etc. Owing to recent development deep learning methods for recognition, performance significantly enhanced datasets. Deep techniques are mainly used recognizing actions images videos comprising Euclidean data. A extension these non-Euclidean data...

It is a strong prerequisite to access source data freely in many existing unsupervised domain adaptation approaches. However, agnostic practical scenarios due the constraints of expensive transmission and privacy protection. Usually, given pre-trained model expected optimize with only unlabeled target data, which termed as source-free adaptation. In this paper, we solve problem from perspective noisy label learning, since can pre-generate for via directly network inference. Under modeling,...

Facial beauty prediction (FBP) aims to develop a machine that automatically makes facial attractiveness assessment. To large extent, the perception of for human is involved with attributes appearance, which provides some significant visual cues FBP. Deep convolution neural networks (CNNs) have shown its power FBP, but filters fixed parameters cannot take full advantage address this problem, we propose an Attribute-aware Convolutional Neural Network (AaNet) modulates main network, adaptively,...

An open research problem in automatic signature verification is the skilled forgery attacks. However, forgeries are very difficult to acquire for representation learning. To tackle this issue, paper proposes learn dynamic representations through ranking synthesized signatures. First, a neuromotor inspired synthesis method proposed synthesize signatures with different distortion levels any template signature. Then, given templates, we construct lightweight one-dimensional convolutional...

Source-free object detection (SFOD) aims to adapt the source detector unlabeled target domain data in absence of data. Most SFOD methods follow same self-training paradigm using mean-teacher (MT) framework where student model is guided by only one single teacher model. However, such can easily fall into a training instability problem that when collapses uncontrollably due shift, also suffers drastic performance degradation. To address this issue, we propose Periodically Exchange...

Human skeleton contains significant information about actions, therefore, it is quite intuitive to incorporate skeletons in human action recognition. resembles a graph where body joints and bones mimic nodes edges. This resemblance of structure the main motivation apply convolutional neural network for Results show that discriminant contribution different not equal actions. Therefore, we propose use attention-joints correspond significantly contributing specific Features corresponding only...

Facial beauty prediction (FBP) aims to automatically assess facial attractiveness consistently with judgements based on human perception. Most of previous methods formulate FBP as a classification, regression or ranking problem machine learning. However, humans not only represent score, but also perceive the relative aesthetics faces. Inspired by this observation, we specific guided information. Specifically, propose general CNN architecture, called R <inline-formula><tex-math...

Handwritten signature verification is a challenging task because signatures of writer may be skillfully imitated by forger. As skilled forgeries are generally difficult to acquire for training, in this paper, we propose deep learning-based dynamic framework, SynSig2Vec, address the forgery attack without training with any forgeries. Specifically, SynSig2Vec consists novel learning-by-synthesis method and 1D convolutional neural network model, called Sig2Vec, representation extraction. The...

Domain generalization (DG) is a fundamental yet very challenging research topic in machine learning. The existing arts mainly focus on learning domain-invariant features with limited source domains static model. Unfortunately, there lack of training-free mechanism to adjust the model when generalized agnostic target domains. To tackle this problem, we develop brand-new DG variant, namely Dynamic Generalization (DDG), which learns twist network parameters adapt data from different...

Facial Beauty Prediction (FBP) is a significant pattern recognition task that aims to achieve consistent facial attractiveness assessment with human perception. Currently, Convolutional Neural Networks (CNNs) have become the mainstream method for FBP. The training objective of most conventional CNNs usually learn static convolution kernels, which, however, makes network quite difficult capture global attentive information, and thus ignores key regions, e.g., eyes, nose. To tackle this...

Self-ensemble adversarial training methods improve model robustness by ensembling models at different epochs, such as weight averaging (WA). However, previous research has shown that self-ensemble defense in (AT) still suffer from robust overfitting, which severely affects the generalization performance. Empirically, late phases of training, AT becomes more overfitting to extent individuals for also and produce anomalous values, causes continue undergo due failure removing anomalies. To...

The purpose of facial beauty prediction (FBP) is to develop a machine that automatically evaluates attractiveness in human perceptual manner. One the essential problem discriminative representation model. Previous methods formulate FBP as specific supervised learning classification, regression, or ranking. We find relative ranking information useful improve regression model FBP. Based on this observation, paper proposes guided by with state-of-the-art Res NeXt structure achieve FBP, and we...

Vanilla unsupervised domain adaptation methods tend to optimize the model with fixed neural architecture, which is not very practical in real-world scenarios since target data usually processed by different resource-limited devices. It therefore of great necessity facilitate architecture across various In this paper, we introduce a simple framework, Slimmable Domain Adaptation, improve cross-domain generalization weight-sharing bank, from models capacities can be sampled accommodate...

This paper proposes a scattering convolutional network with region-aware facial attributes to obtain mid-level representation for beauty prediction (FBP). Different from the previous works that only focus on discriminative prediction, this also considers invariant properties of reduces variances caused by image transformations, such as rotations and translation. The proposed convolution (RegionScatNet) is based deep transforms (ScatNet) integrated texture shape features. It consists three...

In the age of social media, posting attractive mugshots is commonplace, leading to an urgent need for automatic facial beautification techniques. To better meet esthetic preferences users, we devise a customized face task that can retouch adaptively match user-entered target score whilst preserving ID information as much possible. accomplish this task, propose Human Esthetics Guided StyleGAN Inversion method each in embedding space using inversion. This process guided by pre-trained beauty...

Agnostic domain shift is the main reason of model degradation on unknown target domains, which brings an urgent need to develop Domain Generalization (DG). Recent advances at DG use dynamic networks achieve training-free adaptation termed Dynamic (DDG), compensates for lack self-adaptability in static models with fixed weights. The parameters can be decoupled into a and component, are designed learn domain-invariant domain-specific features, respectively. Based existing arts, this work, we...

Facial beauty prediction (FBP) is a significant visual recognition problem to make assessment of facial attractiveness that consistent human perception. To tackle this problem, various data-driven models, especially state-of-the-art deep learning techniques, were introduced, and benchmark dataset become one the essential elements achieve FBP. Previous works have formulated as specific supervised classification, regression or ranking, which indicates FBP intrinsically computation with...

Predicting individual aesthetic preferences holds significant practical applications and academic implications for human society. However, existing studies mainly focus on learning predicting the commonality of facial attractiveness, with little attention given to Personalized Facial Beauty Prediction (PFBP). PFBP aims develop a machine that can adapt only few images rated by each user. In this paper, we formulate task from meta-learning perspective user corresponds meta-task. To address...

Domain generalization (DG) is a fundamental yet very challenging research topic in machine learning. The existing arts mainly focus on learning domain-invariant features with limited source domains static model. Unfortunately, there lack of training-free mechanism to adjust the model when generalized agnostic target domains. To tackle this problem, we develop brand-new DG variant, namely Dynamic Generalization (DDG), which learns twist network parameters adapt data from different...