Cartoon Faces Research Articles

AnyFace++: Deep Multi-Task, Multi-Domain Learning for Efficient Face AI.

Accurate face detection and subsequent localization of facial landmarks are mandatory steps in many computer vision applications, such as emotion recognition, age estimation, and gender identification. Thanks to advancements in deep learning, numerous facial applications have been developed for human faces. However, most have to employ multiple models to accomplish several tasks simultaneously. As a result, they require more memory usage and increased inference time. Also, less attention is paid to other domains, such as animals and cartoon characters. To address these challenges, we propose an input-agnostic face model, AnyFace++, to perform multiple face-related tasks concurrently. The tasks are face detection and prediction of facial landmarks for human, animal, and cartoon faces, including age estimation, gender classification, and emotion recognition for human faces. We trained the model using deep multi-task, multi-domain learning with a heterogeneous cost function. The experimental results demonstrate that AnyFace++ generates outcomes comparable to cutting-edge models designed for specific domains.

Synthesis, Style Editing, and Animation of 3D Cartoon Face

Synthesis, Style Editing, and Animation of 3D Cartoon Face

A Comprehensive Framework for Face age Synthesis, Voice Transformation, and Cartoon Generation

This paper presents a comprehensive framework addressing age-invariant face recognition (AIFR), face age synthesis (FAS), voice transformation, and avatar generation. Traditional AIFR techniques focus on minimizing age-related variations in face recognition but lack visual interpretability, while FAS methods aim to synthesize faces of different ages but often compromise recognition due to artifacts. To address these limitations, we propose MTLFace, a multi-task learning framework that simultaneously handles AIFR and FAS tasks. MTLFace employs attention-based feature decomposition to separate identity and age-related features spatially, improving interpretability. Additionally, we introduce an identity conditional module for fine-grained face age synthesis, enhancing the naturalness of synthesized faces. Unlike conventional methods that achieve age group-level synthesis, our identity conditional module enables identity-level synthesis, resulting in smoother age transitions and preserving individual facial characteristics. Leveraging high-quality synthesized faces, we enhance AIFR performance via selective fine-tuning, where synthesized faces are used to augment training data, leading to improved robustness against age variations. Furthermore, we contribute a large cross-age face dataset with annotations, facilitating research in age-related tasks. In addition to AIFR and FAS, we explore limitations in existing methodologies for voice transformation and propose advancements to improve speech quality and address current challenges. Lastly, we introduce a novel approach for cartoon face generation, utilizing component-based facial feature extraction and template matching to produce diverse and stylistic cartoon faces. Experimental results across various benchmarks demonstrate the effectiveness of our approaches in achieving superior performance in face recognition, voice transformation, and cartoon face generation, thereby contributing to the advancement of multimodal synthesis technologies with potential applications in entertainment, virtual communication, and assistive technologies

Convolutional neural network for classifying cartoon images augmented by DCGAN

Convolutional Neural Network (CNN) tend to have better results on large data sets and poor performance on small data sets, so the data augmentation is crucial for a CNN to get better performance based on the dataset with limited size. In this paper, Deep Convolution Generative Adversarial Network (DCGAN) was used to augment data to make the AlexNet perform better on an image classification task with small data sets. AlexNet was trained on a small anime face training set with only 160 samples to determine whether the anime face was male or female, and then tested its accuracy on a test set with 240 sample. Then, a pre-trained DCGAN was transferred to train on the male and female training sets respectively. And 2 DCGANs were obtained, one could generate male cartoon faces and another could generate female cartoon faces. The images generated by DCGANs were put in train set, which was used to train AlexNet again and the result was recorded. Other data augmentation methods such as cutout, cutmix and Noise Injection were compared as well. Finally, it is found that AlexNet has the best performance when using the DCGAN augmentation method, which can significantly improve the verification accuracy of the model.

Machine learning for distinguishing saudi children with and without autism via eye-tracking data

BackgroundDespite the prevalence of Autism Spectrum Disorder (ASD) globally, there’s a knowledge gap pertaining to autism in Arabic nations. Recognizing the need for validated biomarkers for ASD, our study leverages eye-tracking technology to understand gaze patterns associated with ASD, focusing on joint attention (JA) and atypical gaze patterns during face perception. While previous studies typically evaluate a single eye-tracking metric, our research combines multiple metrics to capture the multidimensional nature of autism, focusing on dwell times on eyes, left facial side, and joint attention.MethodsWe recorded data from 104 participants (41 neurotypical, mean age: 8.21 ± 4.12 years; 63 with ASD, mean age 8 ± 3.89 years). The data collection consisted of a series of visual stimuli of cartoon faces of humans and animals, presented to the participants in a controlled environment. During each stimulus, the eye movements of the participants were recorded and analyzed, extracting metrics such as time to first fixation and dwell time. We then used these data to train a number of machine learning classification algorithms, to determine if these biomarkers can be used to diagnose ASD.ResultsWe found no significant difference in eye-dwell time between autistic and control groups on human or animal eyes. However, autistic individuals focused less on the left side of both human and animal faces, indicating reduced left visual field (LVF) bias. They also showed slower response times and shorter dwell times on congruent objects during joint attention (JA) tasks, indicating diminished reflexive joint attention. No significant difference was found in time spent on incongruent objects during JA tasks. These results suggest potential eye-tracking biomarkers for autism. The best-performing algorithm was the random forest one, which achieved accuracy = 0.76 ± 0.08, precision = 0.78 ± 0.13, recall = 0.84 ± 0.07, and F1 = 0.80 ± 0.09.ConclusionsAlthough the autism group displayed notable differences in reflexive joint attention and left visual field bias, the dwell time on eyes was not significantly different. Nevertheless, the machine algorithm model trained on these data proved effective at diagnosing ASD, showing the potential of these biomarkers. Our study shows promising results and opens up potential for further exploration in this under-researched geographical context.

Is the traditional placement of stainless steel crowns in primary teeth more painful than the non-preparation Hall technique?

Is the anxiety level and pain experience in young children the same for the traditional method of stainless crowns placement that might require local anesthetic and excavation of decay compared to the Hall technique that requires no preparation and no anesthetic injections? A randomized controlled study of 2 groups (43 and 54) children 4-8 years old with proximal decay in primary first molars. Each child had one tooth restored with the Hall technique (HT) and the second tooth received a conventional stainless crown (CT) (with or without local anesthesia, as needed) and all crowns were cemented with a GIC resin type cement less than a month of each other. The order of the placement was reversed with the allocation randomly concealed. A visual Wong-Baker Pain Scale with six cartoon faces was used to record pain levels. Data were analyzed using SPSS (version 25). After computing univariate descriptive statistics, differences among proportions were examined using cross-tabulations and χ2 tests (α = 0.05). Analysis of variance was conducted for parametric analysis, and the Wilcoxon signed-rank test was used for within-group, and the Mann-Whitney U-test was used for between-group, nonparametric analysis. The pain scores did not differ by treatment type. The majority of children (58.3%) reported low procedural pain for both treatments leaving 47.7% reporting moderate to high procedural pain. Using local anesthetic with CT did not change the results. The pain scores were similar for both procedures. HT has been recommended as less traumatic and but this might not be true for all children.

Abstraction-perception preserving cartoon face synthesis

Abstraction-perception preserving cartoon face synthesis

Visual consequent stimulus complexity affects performance in audiovisual associative learning

In associative learning (AL), cues and/or outcome events are coupled together. AL is typically tested in visual learning paradigms. Recently, our group developed various AL tests based on the Rutgers Acquired Equivalence Test (RAET), both visual and audiovisual, keeping the structure and logic of RAET but with different stimuli. In this study, 55 volunteers were tested in two of our audiovisual tests, SoundFace (SF) and SoundPolygon (SP). The antecedent stimuli in both tests are sounds, and the consequent stimuli are images. The consequents in SF are cartoon faces, while in SP, they are simple geometric shapes. The aim was to test how the complexity of the applied consequent stimuli influences performance regarding the various aspects of learning the tests assess (stimulus pair learning, retrieval, and generalization of the previously learned associations to new but predictable stimulus pairs). In SP, behavioral performance was significantly poorer than in SF, and the reaction times were significantly longer, for all phases of the test. The results suggest that audiovisual associative learning is significantly influenced by the complexity of the consequent stimuli.

A Domain-Guided Model for Facial Cartoonlization

Dear Editor, This work investigates the issue of facial cartoonlization under the condition of lacking training data. We propose a domain-guided model (DGM) to realize facial cartoonlization for different kinds of faces. It includes two parts: 1) a domain-guided model that contains four different interface networks and can embed an image from a facial domain to a cartoon domain independently; and 2) a one-to-one tutoring strategy that uses a sub-model as a teacher to train other interface networks and can yield fine-grained cartoon faces. Extensive qualitative and quantitative experimental results validate our proposed method, and show that DGM can yield fine-translated results for different kinds of faces and outperforms the state of the art.

Attentional biases toward real images and drawings of negative faces

The allocation of attention is affected by internal emotional states, such as anxiety and depression. The attention captured by real images of negative faces can be quantified by emotional probe tasks. The present study investigated whether attentional bias toward drawings of negative faces (line drawings and cartoon faces) differs from that of real faces. Non-clinical university students indicated their levels of anxiety and depression via self-report questionnaires, and completed a probe discrimination task under three face image conditions in a between-participants design. Significant correlations were found between bias scores and scores on the self-reported BDI-II under the real face condition. However, two types of face drawings were only weakly correlated with self-report scores. In our probe task to investigate attentional bias to facial stimuli in nonclinical adults, the strength of the relationship between depression and attentional bias to negative face was stronger for real faces than for face drawings.

CAMGAN: Combining attention mechanism generative adversarial networks for cartoon face style transfer

In this paper, we presents an apporch for real-world human face close-up images cartoonization. We use generative adversarial network combined with an attention mechanism to convert real-world face pictures and cartoon-style images as unpaired data sets. At present, the image-to-image translation model has been able to successfully transfer style and content. However, some problems still exist in the task of cartoonizing human faces:Hunman face has many details, and the content of the image is easy to lose details after the image is translated. the quality of the image generated by the model is defective. The model in this paper uses the generative adversarial network combined with the attention mechanism, and proposes a new generative adversarial network combined with the attention mechanism to deal with these problems. The channel attention mechanism is embedded between the upper and lower sampling layers of the generator network, to avoid increasing the complexity of the model while conveying the complete details of the underlying information. After comparing the experimental results of FID, PSNR, MSE three indicators and the size of the model parameters, the new model network proposed in this paper avoids the complexity of the model while achieving a good balance in the conversion task of style and content.

Human-robot interaction: the impact of robotic aesthetics on anticipated human trust.

BackgroundHuman senses have evolved to recognise sensory cues. Beyond our perception, they play an integral role in our emotional processing, learning, and interpretation. They are what help us to sculpt our everyday experiences and can be triggered by aesthetics to form the foundations of our interactions with each other and our surroundings. In terms of Human-Robot Interaction (HRI), robots have the possibility to interact with both people and environments given their senses. They can offer the attributes of human characteristics, which in turn can make the interchange with technology a more appealing and admissible experience. However, for many reasons, people still do not seem to trust and accept robots. Trust is expressed as a person’s ability to accept the potential risks associated with participating alongside an entity such as a robot. Whilst trust is an important factor in building relationships with robots, the presence of uncertainties can add an additional dimension to the decision to trust a robot. In order to begin to understand how to build trust with robots and reverse the negative ideology, this paper examines the influences of aesthetic design techniques on the human ability to trust robots.MethodThis paper explores the potential that robots have unique opportunities to improve their facilities for empathy, emotion, and social awareness beyond their more cognitive functionalities. Through conducting an online questionnaire distributed globally, we explored participants ability and acceptance in trusting the Canbot U03 robot. Participants were presented with a range of visual questions which manipulated the robot’s facial screen and asked whether or not they would trust the robot. A selection of questions aimed at putting participants in situations where they were required to establish whether or not to trust a robot’s responses based solely on the visual appearance. We accomplished this by manipulating different design elements of the robots facial and chest screens, which influenced the human-robot interaction.ResultsWe found that certain facial aesthetics seem to be more trustworthy than others, such as a cartoon face versus a human face, and that certain visual variables (i.e., blur) afforded uncertainty more than others. Consequentially, this paper reports that participant’s uncertainties of the visualisations greatly influenced their willingness to accept and trust the robot. The results of introducing certain anthropomorphic characteristics emphasised the participants embrace of the uncanny valley theory, where pushing the degree of human likeness introduced a thin line between participants accepting robots and not. By understanding what manipulation of design elements created the aesthetic effect that triggered the affective processes, this paper further enriches our knowledge of how we might design for certain emotions, feelings, and ultimately more socially acceptable and trusting robotic experiences.

Graph Jigsaw Learning for Cartoon Face Recognition.

Cartoon face recognition is challenging as they typically have smooth color regions and emphasized edges, the key to recognizing cartoon faces is to precisely perceive their sparse and critical shape patterns. However, it is quite difficult to learn a shape-oriented representation for cartoon face recognition with convolutional neural networks (CNNs). To mitigate this issue, we propose the GraphJigsaw that constructs jigsaw puzzles at various stages in the classification network and solves the puzzles with the graph convolutional network (GCN) in a progressive manner. Solving the puzzles requires the model to spot the shape patterns of the cartoon faces as the texture information is quite limited. The key idea of GraphJigsaw is constructing a jigsaw puzzle by randomly shuffling the intermediate convolutional feature maps in the spatial dimension and exploiting the GCN to reason and recover the correct layout of the jigsaw fragments in a self-supervised manner. The proposed GraphJigsaw avoids training the classification model with the deconstructed images that would introduce noisy patterns and are harmful for the final classification. Specially, GraphJigsaw can be incorporated at various stages in a top-down manner within the classification model, which facilitates propagating the learned shape patterns gradually. GraphJigsaw does not rely on any extra manual annotation during the training process and incorporates no extra computation burden at inference time. Both quantitative and qualitative experimental results have verified the feasibility of our proposed GraphJigsaw, which consistently outperforms other face recognition or jigsaw-based methods on two popular cartoon face datasets with considerable improvements.

Morphing Task: The Emotion Recognition Process in Children with Attention Deficit Hyperactivity Disorder and Autism Spectrum Disorder.

Recognizing a person’s identity is a fundamental social ability; facial expressions, in particular, are extremely important in social cognition. Individuals affected by autism spectrum disorder (ASD) and attention deficit hyperactivity disorder (ADHD) display impairment in the recognition of emotions and, consequently, in recognizing expressions related to emotions, and even their identity. The aim of our study was to compare the performance of participants with ADHD, ASD, and typical development (TD) with regard to both accuracy and speed in the morphing task and to determine whether the use of pictures of digitized cartoon faces could significantly facilitate the process of emotion recognition in ASD patients (particularly for disgust). This study investigated the emotion recognition process through the use of dynamic pictures (human faces vs. cartoon faces) created with the morphing technique in three pediatric populations (7–12 years old): ADHD patients, ASD patients, and an age-matched control sample (TD). The Chi-square test was used to compare response latency and accuracy between the three groups in order to determine if there were statistically significant differences (p < 0.05) in the recognition of basic emotions. The results demonstrated a faster response time in neurotypical children compared to ASD and ADHD children, with ADHD participants performing better than ASD participants on the same task. The overall accuracy parameter between the ADHD and ASD groups did not significantly differ.

The Influence of Key Facial Features on Recognition of Emotion in Cartoon Faces.

Cartoon faces are widely used in social media, animation production, and social robots because of their attractive ability to convey different emotional information. Despite their popular applications, the mechanisms of recognizing emotional expressions in cartoon faces are still unclear. Therefore, three experiments were conducted in this study to systematically explore a recognition process for emotional cartoon expressions (happy, sad, and neutral) and to examine the influence of key facial features (mouth, eyes, and eyebrows) on emotion recognition. Across the experiments, three presentation conditions were employed: (1) a full face; (2) individual feature only (with two other features concealed); and (3) one feature concealed with two other features presented. The cartoon face images used in this study were converted from a set of real faces acted by Chinese posers, and the observers were Chinese. The results show that happy cartoon expressions were recognized more accurately than neutral and sad expressions, which was consistent with the happiness recognition advantage revealed in real face studies. Compared with real facial expressions, sad cartoon expressions were perceived as sadder, and happy cartoon expressions were perceived as less happy, regardless of whether full-face or single facial features were viewed. For cartoon faces, the mouth was demonstrated to be a feature that is sufficient and necessary for the recognition of happiness, and the eyebrows were sufficient and necessary for the recognition of sadness. This study helps to clarify the perception mechanism underlying emotion recognition in cartoon faces and sheds some light on directions for future research on intelligent human-computer interactions.

My own face looks larger than yours: A self-induced illusory size perception

Size perception of visual objects is highly context dependent. Here we report a novel perceptual size illusion that the self-face, being a unique and distinctive self-referential stimulus, can enlarge its perceived size. By using a size discrimination paradigm, we found that the self-face was perceived as significantly larger than the other-face of the same size. This size overestimation effect was not due to the familiarity of the self-face, since it could be still observed when the self-face was directly compared with a famous face. More crucially, such illusion effect could be extended to a new cartoon face that was transiently associated with one's own face and could also exert further contextual influences on visual size perception of other objects. These findings together highlight the role of self-awareness in visual size perception and point to a special mechanism of size perception tuned to self-referential information.

Effects of social information on duration perception by different mechanisms in sub- and supra-second range: Evidence from face features.

Human faces convey abundant social information. The influence of social information on duration estimation was examined by comparing a cartoon face (CF; social) condition with geometric plane figures, or CFs with internal elements disturbed from their normal position (non-social) condition, and also by comparing a human portrait facing the front (more social interaction) condition with the portrait facing the back (less social interaction) condition. Results showed that in the supra-second range, both CFs and human portraits facing the front led to an underestimation compared with non-social or less social information conditions. In the sub-second range, only portraits facing the front were overestimated compared with those facing the back and with the CFs. These results suggest that in the millisecond range, the biological information conveyed by faces may lead to higher arousal, which may result in accelerated clock speed and overestimation. In the supra-second range, it is possible that attention was diverted by social information, resulting in duration underestimation. Taken together, our results suggest that the different mechanisms were involved in duration perception of social information conveyed by face feature in the sub- and supra-second range.

Single-Option P300-BCI Performance Is Affected by Visual Stimulation Conditions.

The P300 paradigm is one of the most promising techniques for its robustness and reliability in Brain-Computer Interface (BCI) applications, but it is not exempt from shortcomings. The present work studied single-trial classification effectiveness in distinguishing between target and non-target responses considering two conditions of visual stimulation and the variation of the number of symbols presented to the user in a single-option visual frame. In addition, we also investigated the relationship between the classification results of target and non-target events when training and testing the machine-learning model with datasets containing different stimulation conditions and different number of symbols. To this end, we designed a P300 experimental protocol considering, as conditions of stimulation: the color highlighting or the superimposing of a cartoon face and from four to nine options. These experiments were carried out with 19 healthy subjects in 3 sessions. The results showed that the Event-Related Potentials (ERP) responses and the classification accuracy are stronger with cartoon faces as stimulus type and similar irrespective of the amount of options. In addition, the classification performance is reduced when using datasets with different type of stimulus, but it is similar when using datasets with different the number of symbols. These results have a special connotation for the design of systems, in which it is intended to elicit higher levels of evoked potentials and, at the same time, optimize training time.

自闭症谱系障碍儿童对真人和卡通的面孔情绪加工

自闭症谱系障碍儿童对真人和卡通的面孔情绪加工

Automatic Processing Advantage of Cartoon Face in Internet Gaming Disorder: Evidence From P100, N170, P200, and MMN.

Individuals with Internet gaming disorder (IGD) show deficits in face processing due to long-term Internet-game social activities based on cartoon faces in the popular online game "Strike of Kings." However, the abnormal neurocognitive mechanism of face recognition and processing in individuals with IGD has not been systematically explored. This study used event-related potential (ERP) methods and the reversed deviant-standard oddball paradigm to comprehensively compare four ERP components, namely, P100, N170, P200, and mismatch negativity (MMN), induced in the unconscious and automatic processing of realistic and cartoon faces in individuals with IGD. Results showed that, with respect to cartoon faces, individuals with IGD exhibited not only P100, P200 and MMN enhancements but also the absence of the N170 dominance effect in the left hemisphere. Our results also demonstrated that individuals with IGD had the advantages of early automatic perception of cartoon faces and automatic detection of changes in "cartoon" features. This study enhances our understanding of the mechanism of IGD from the neurocognitive perspective and provides candidate electrophysiological indicators for the clinical diagnosis of IGD.