Facial Information Research Articles

Comparing internal representations of facial expression kinematics between autistic and non-autistic adults.

Recent developments suggest that autistic individuals require dynamic angry expressions to have a higher speed in order for them to be successfully identified. Therefore, it is plausible that autistic individuals do not have a 'deficit' in angry expression recognition, but rather their internal representation of these expressions is characterised by very high-speed movement. In this study, matched groups of autistic and non-autistic adults completed a novel emotion-based task which employed dynamic displays of happy, angry and sad point light facial (PLF) expressions. On each trial, participants moved a slider to manipulate the speed of a PLF stimulus until it moved at a speed that, in their 'mind's eye', was typical of happy, angry or sad expressions. Participants were shown three different types of PLFs-those showing the full-face, only the eye region, and only the mouth region, wherein the latter two were included to test whether differences in facial information sampling underpinned any dissimilarities in speed attributions. Across both groups, participants attributed the highest speeds to angry, then happy, then sad, facial motion. Participants increased the speed of angry and happy expressions by 41% and 27% respectively and decreased the speed of sad expressions by 18%. This suggests that participants have 'caricatured' internal representations of emotion, wherein emotion-related kinematic cues are over-emphasised. There were no differences between autistic and non-autistic individuals in the speeds attributed to full-face and partial-face angry, happy and sad expressions respectively. Consequently, we find no evidence that autistic adults possess atypically fast internal representations of anger.

Refining Orthognathic Surgery Results by Synchronous Cheek Fat Compartment Augmentation with Fat Grafting in Adult Females with Class III Skeletal Profiles.

Patients with maxillomandibular disharmony may present with a flat to concave midface. The effects of orthognathic surgery concomitant with midface fat grafting on facial appearance and midface volumetric and positional change have not formally been assessed to date. The authors' approach for synchronous orthognathic surgery and fat grafting is described and evaluated. Adult female patients (n = 20) who underwent synchronous two-jaw orthognathic surgery and cheek-specific fat grafting (1.9 ± 0.6 cm3 per side) for correction of skeletal class III deformity and anteromedial cheek deficiency were prospectively included. Preoperative and postoperative photographs were appraised by 42 blinded raters using facial appearance scales for beauty, attractiveness, and pleasantness parameters. The three-dimensional midface soft-tissue volume change and postoperative cheek mass position were computed. Facial imaging data from gender-, ethnic-, and facial pattern-matched adult patients (n = 20) who underwent isolated two-jaw orthognathic surgery (n = 20) were included for comparison. The three-dimensional facial norms database-derived cheek mass position information (2.19 ± 1.31mm) was also adopted for analysis. Patients treated with the synchronous procedure had significantly (p < 0.001) increased facial appearance-related perception change for beauty (2.9 ± 1.6), attractiveness (2.8 ± 1.8), and pleasantness (3.0 ± 1.5) parameters, three-dimensional midface volume change (1.8 ± 0.5 cm3), and postoperative cheek mass position (2.16 ± 0.47 mm) in comparison with those treated with the isolated procedure (2.0 ± 1.5, 1.9 ± 1.6, 2.3 ± 1.6, 0.6 ± 0.2 cm3, and 1.84 ± 0.43 mm, respectively). Healthy female individuals had similar and larger cheek mass position than patients treated with synchronous (p > 0.05) and isolated (p < 0.001) procedures, respectively. Synchronous orthognathic surgery and check-specific fat grafting resulted in superior enhancement of facial appearance and midface volume and position compared with isolated orthognathic surgery. Therapeutic, III.

A Meta-Analysis of Privacy

Facial recognition systems use advanced computing to capture facial information and compare the same with proprietary databases for validation. The emergence of data capturing intermediaries and open access image repositories have compounded the need for a holistic perspective for handling the privacy and security challenges associated with FRS. The study presents the results of a bibliometric analysis conducted on the topic of privacy, ethical and security aspects of FRS. This study presents the level of academic discussion on the topic using bibliometric performance analysis. The results of the bibliographic coupling analysis to identify the research hotspots are also presented. The results also include the systematic literature review of 148 publications that are distributed across seven themes. Both the bibliometric and systematic analysis showed that privacy and security in FRS requires a holistic perspective that cuts across privacy, ethical, security, legal, policy and technological aspects.

Deepfake Detection Method Based on Cross-Domain Fusion

In recent years, despite its wide use in various fields, deepfake has been abused to generate hazardous contents such as fake movies, rumors, and fake news by manipulating or replacing facial information of the original sources and, thus, exerts huge security threats to the society. Facing the continuous evolution of deepfake, research on active detection and prevention technology becomes particularly important. In this paper, we propose a new deepfake detection method based on cross-domain fusion, which, on the basis of traditional spatial domain features, realizes the fusion of cross-domain image features by introducing edge geometric features of the frequency domain and, therefore, achieves considerable improvements on classification accuracy. Further evaluations of this method have been performed on publicly deepfake datasets, and the results show that our method is effective particularly on the Meso-4 DeepFake Database.

Diagnostic Features for Human Categorisation of Adult and Child Faces.

Knowing how humans differentiate children from adults has useful implications in many areas of both forensic and cognitive psychology. Yet, how we extract age from faces has been surprisingly underexplored in both disciplines. Here, we used a novel data-driven experimental technique to objectively measure the facial features human observers use to categorise child and adult faces. Relying on more than 35,000 trials, we used a reverse correlation technique that enabled us to reveal how specific features which are known to be important in face-perception – position, spatial-frequency (SF), and orientation – are associated with accurate child and adult discrimination. This showed that human observers relied on evidence in the nasal bone and eyebrow area for accurate adult categorisation, while they relied on the eye and jawline area to accurately categorise child faces. For orientation structure, only facial information of vertical orientation was linked to face-adult categorisation, while features of horizontal and, to a lesser extent oblique orientations, were more diagnostic of a child face. Finally, we found that SF diagnosticity showed a U-shaped pattern for face-age categorisation, with information in low and high SFs being diagnostic of child faces, and mid SFs being diagnostic of adult faces. Through this first characterisation of the facial features of face-age categorisation, we show that important information found in psychophysical studies of face-perception in general (i.e., the eye area, horizontals, and mid-level SFs) is crucial to the practical context of face-age categorisation, and present data-driven procedures through which face-age classification training could be implemented for real-world challenges.

Multimodal Emotion Recognition on RAVDESS Dataset Using Transfer Learning.

Emotion Recognition is attracting the attention of the research community due to the multiple areas where it can be applied, such as in healthcare or in road safety systems. In this paper, we propose a multimodal emotion recognition system that relies on speech and facial information. For the speech-based modality, we evaluated several transfer-learning techniques, more specifically, embedding extraction and Fine-Tuning. The best accuracy results were achieved when we fine-tuned the CNN-14 of the PANNs framework, confirming that the training was more robust when it did not start from scratch and the tasks were similar. Regarding the facial emotion recognizers, we propose a framework that consists of a pre-trained Spatial Transformer Network on saliency maps and facial images followed by a bi-LSTM with an attention mechanism. The error analysis reported that the frame-based systems could present some problems when they were used directly to solve a video-based task despite the domain adaptation, which opens a new line of research to discover new ways to correct this mismatch and take advantage of the embedded knowledge of these pre-trained models. Finally, from the combination of these two modalities with a late fusion strategy, we achieved 80.08% accuracy on the RAVDESS dataset on a subject-wise 5-CV evaluation, classifying eight emotions. The results revealed that these modalities carry relevant information to detect users’ emotional state and their combination enables improvement of system performance.

Effective Face Detector Based on YOLOv5 and Superresolution Reconstruction.

The application of face detection and recognition technology in security monitoring systems has made a huge contribution to public security. Face detection is an essential first step in many face analysis systems. In complex scenes, the accuracy of face detection would be limited because of the missing and false detection of small faces, due to image quality, face scale, light, and other factors. In this paper, a two-level face detection model called SR-YOLOv5 is proposed to address some problems of dense small faces in actual scenarios. The research first optimized the backbone and loss function of YOLOv5, which is aimed at achieving better performance in terms of mean average precision (mAP) and speed. Then, to improve face detection in blurred scenes or low-resolution situations, we integrated image superresolution technology on the detection head. In addition, some representative deep-learning algorithm based on face detection is discussed by grouping them into a few major categories, and the popular face detection benchmarks are enumerated in detail. Finally, the wider face dataset is used to train and test the SR-YOLOv5 model. Compared with multitask convolutional neural network (MTCNN), Contextual Multi-Scale Region-based CNN (CMS-RCNN), Finding Tiny Faces (HR), Single Shot Scale-invariant Face Detector (S3FD), and TinaFace algorithms, it is verified that the proposed model has higher detection precision, which is 0.7%, 0.6%, and 2.9% higher than the top one. SR-YOLOv5 can effectively use face information to accurately detect hard-to-detect face targets in complex scenes.

Deep-Learning-Based Stress Recognition with Spatial-Temporal Facial Information.

In recent times, as interest in stress control has increased, many studies on stress recognition have been conducted. Several studies have been based on physiological signals, but the disadvantage of this strategy is that it requires physiological-signal-acquisition devices. Another strategy employs facial-image-based stress-recognition methods, which do not require devices, but predominantly use handcrafted features. However, such features have low discriminating power. We propose a deep-learning-based stress-recognition method using facial images to address these challenges. Given that deep-learning methods require extensive data, we constructed a large-capacity image database for stress recognition. Furthermore, we used temporal attention, which assigns a high weight to frames that are highly related to stress, as well as spatial attention, which assigns a high weight to regions that are highly related to stress. By adding a network that inputs the facial landmark information closely related to stress, we supplemented the network that receives only facial images as the input. Experimental results on our newly constructed database indicated that the proposed method outperforms contemporary deep-learning-based recognition methods.

Spontaneous Visual Preference for Face-Like Stimuli Is Impaired in Newly-Hatched Domestic Chicks Exposed to Valproic Acid During Embryogenesis.

Faces convey a great amount of socially relevant information related to emotional and mental states, identity and intention. Processing of face information is a key mechanism for social and cognitive development, such that newborn babies are already tuned to recognize and orient to faces and simple schematic face-like patterns since the first hours of life. Similar to neonates, also non-human primates and domestic chicks have been shown to express orienting responses to faces and schematic face-like patterns. More importantly, existing studies have hypothesized that early disturbances of these mechanisms represent one of the earliest biomarker of social deficits in autism spectrum disorders (ASD). We used VPA exposure to induce neurodevelopmental changes associated with ASD in domestic chicks and tested whether VPA could impact the expression of the animals’ approach responses to schematic face-like stimuli. We found that VPA impairs the chicks’ preference responses to these social stimuli. Based on the results shown here and on previous studies, we propose the domestic chick as animal model to investigate the biological mechanisms underlying face processing deficits in ASD.

Lightweight facial expression recognition method based on attention mechanism and key region fusion

To improve the facial expression recognition accuracy in resource-constrained and real-time application equipment such as mobile and embedded devices, a lightweight method for facial expression recognition is proposed based on attention mechanism and key regions fusion. To reduce the computation complexity, a lightweight convolutional neural network, mini_Xception, is used as the basic expression recognition model for expression classification. The attention mechanism is introduced to enhance the learning of the important features of the whole face. Then a parameter is introduced to locate the key regions and construct key region models. Finally, to realize the complementarity of models and learn more comprehensive features, the whole facial expression recognition model is fused with the key region models. The proposed method can capture and utilize the important facial expression information in related regions displayed through class activation mapping visualization. The experimental results on JAFFE, CK+ datasets, and a real scene dataset verify the effectiveness of the proposed method.

Face to purchase: Predicting consumer choices with structured facial and behavioral traits embedding

Predicting consumers’ purchasing behaviors is critical for targeted advertisement and sales promotion in e-commerce. Human faces are an invaluable source of information for gaining insights into consumer personality and behavioral traits. However, consumer’s faces are largely unexplored in previous research, and the existing face-related studies focus on high-level features such as personality traits while neglecting the business significance of learning from facial data. We propose to predict consumers’ purchases based on their facial features and purchasing histories. We design a semi-supervised model based on a hierarchical embedding network to extract high-level features of consumers and to predict the top-N purchase destinations of a consumer. Our experimental results on a real-world dataset demonstrate the positive effect of incorporating facial information in predicting consumers’ purchasing behaviors.

Driver Gaze Zone Estimation via Head Pose Fusion Assisted Supervision and Eye Region Weighted Encoding

Driver gaze zone estimation is an important task in Advanced Driver Assistance Systems (ADAS), which suffers difficulties including head pose, capture direction, glass occlusion, and real-time requirement, etc. Most previous methods combine face modalities and head pose using concat process, which may result in over-fitting due to the unbalanced dimension. Focusing on gaze zone estimation problems, we propose the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Head Pose Fusion Assisted supervision &amp; Eye Region Weighted Encoding</i> ( <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">HP-ERW</b> ) structure, which fuses head pose attribute and face modalities together through spatial attention and Kronecker product mechanisms. Firstly, we introduce a pre-processing module dealing with head pose and face information, with the purpose of extracting input vectors and improving the fusion speed of the HP-ERW structure. Secondly, an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Eye Region Weighted Encoding Network</i> (ERW-Net) based on spatial attention is proposed to strengthen the networks perception ability for encoding features. Finally, we propose a dual-channel <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Head Pose Fusion Network</i> ( <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">HP-Net</b> ) based on the Kronecker product mechanism, with the purpose of fusing head pose and improving the estimation accuracy. Experiments show that the HP-ERW outperforms compared existing methods on several public datasets. The designed ADAS using the proposed method achieves 23.5 fps real-time application with small memory requirement of 4,884 KB.

Face Adaptation-Investigating Nonconfigural Saturation Alterations.

Recognizing familiar faces requires a comparison of the incoming perceptual information with mental face representations stored in memory. Mounting evidence indicates that these representations adapt quickly to recently perceived facial changes. This becomes apparent in face adaptation studies where exposure to a strongly manipulated face alters the perception of subsequent face stimuli: original, non-manipulated face images then appear to be manipulated, while images similar to the adaptor are perceived as “normal.” The face adaptation paradigm serves as a good tool for investigating the information stored in facial memory. So far, most of the face adaptation studies focused on configural (second-order relationship) face information, mainly neglecting non-configural face information (i.e., that does not affect spatial face relations), such as color, although several (non-adaptation) studies were able to demonstrate the importance of color information in face perception and identification. The present study therefore focuses on adaptation effects on saturation color information and compares the results with previous findings on brightness. The study reveals differences in the effect pattern and robustness, indicating that adaptation effects vary considerably even within the same class of non-configural face information.

Generative Facial Prior for Large-Factor Blind Face Super-Resolution

Most existing face super-resolution (SR) methods are developed based on an assumption that the degradation is fixed and known (e.g., bicubic down sampling). However, these methods suffer a severe performance drop in various unknown degradations in real-world applications. Previous methods usually rely on facial priors, such as facial geometry prior or reference prior, to restore realistic face details. Nevertheless, low-quality inputs cannot provide accurate geometric priors while high-quality references are often unavailable, which limits the use of face super-resolution in real-world scenes. In this work, we propose GPLSR which used the rich priors encapsulated in the pre-trained face GAN network to perform blind face super-resolution. This generative facial priori is introduced into the face super-resolution process through channel squeeze-and-excitation spatial feature transformation layer (SE-SFT), which makes our method achieve a good balance between realness and fidelity. Moreover, GPLSR can restores facial details with single forward pass because of powerful generative facial prior information. Extensive experiment shows that when the magnification factor is 16, this method achieves better performance than existing techniques in both synthetic and real datasets.

Avatars for Teleconsultation: Effects of Avatar Embodiment Techniques on User Perception in 3D Asymmetric Telepresence.

A 3D Telepresence system allows users to interact with each other in a virtual, mixed, or augmented reality (VR, MR, AR) environment, creating a shared space for collaboration and communication. There are two main methods for representing users within these 3D environments. Users can be represented either as point cloud reconstruction-based avatars that resemble a physical user or as virtual character-based avatars controlled by tracking the users' body motion. This work compares both techniques to identify the differences between user representations and their fit in the reconstructed environments regarding the perceived presence, uncanny valley factors, and behavior impression. Our study uses an asymmetric VR/AR teleconsultation system that allows a remote user to join a local scene using VR. The local user observes the remote user with an AR head-mounted display, leading to facial occlusions in the 3D reconstruction. Participants perform a warm-up interaction task followed by a goal-directed collaborative puzzle task, pursuing a common goal. The local user was represented either as a point cloud reconstruction or as a virtual character-based avatar, in which case the point cloud reconstruction of the local user was masked. Our results show that the point cloud reconstruction-based avatar was superior to the virtual character avatar regarding perceived co-presence, social presence, behavioral impression, and humanness. Further, we found that the task type partly affected the perception. The point cloud reconstruction-based approach led to higher usability ratings, while objective performance measures showed no significant difference. We conclude that despite partly missing facial information, the point cloud-based reconstruction resulted in better conveyance of the user behavior and a more coherent fit into the simulation context.

Skin coloration is a culturally-specific cue for attractiveness, healthiness, and youthfulness in observers of Chinese and western European descent.

Facial skin coloration signals information about an individual and plays an important role in social interactions and mate choice, due its putative association with health, attractiveness, and age. Whether skin coloration as an evolutionary significant cue is universal or specific to a particular culture is unclear and current evidence on the universality of skin color as a cue to health and attractiveness are equivocal. The current study used 80 calibrated, high-resolution, non-manipulated images of real human faces, either of Chinese or western European descent, which were rated in terms of attractiveness, healthiness, and perceived age by 44 observers, 22 western European (13 male; mean age ± SD = 24.27 ± 5.30) and 22 Chinese (7 male; mean age ± SD = 26.05 ± 3.96) observers. To elucidate the associations between skin coloration and these perceptual ratings and whether these associations are modulated by observer or image ethnicity, a linear mixed-effect model was setup with skin lightness (L*; CIELAB), redness (a*) and yellowness (b*), observer and image ethnicity as independent variables and perceived attractiveness, healthiness, and estimated age as dependent variables. We found robust positive associations between facial skin lightness (L*) and attractiveness, healthiness, and youthfulness, but only when Chinese observers judge facial images of their own ethnicity. Observers of European descent, on the other hand, associated an increase in yellowness(b*) with greater attractiveness and healthiness in Chinese facial images. We find no evidence that facial redness is positively associated with these attributes; instead, an increase in redness (a*) is associated with an increase in the estimated age of European facial images. We conclude that observers of both ethnicities make use of skin color and lightness to rate attractiveness, healthiness, and perceived age, but to a lesser degree than previously thought. Furthermore, these coloration cues are not universal and are utilized differently within the Chinese and western European ethnic groups. Our study adds to the growing body of work demonstrating the importance of skin color manipulations within an evolutionary meaningful parameter space, ideally using realistic skin models based on physical parameters.

Spatially Adjacent Regions in Posterior Cingulate Cortex Represent Familiar Faces at Different Levels of Complexity.

Extensive research has shown that perceptual information of faces is processed in a network of hierarchically-organized areas within ventral temporal cortex. For familiar and famous faces, perceptual processing of faces is normally accompanied by extraction of semantic knowledge about the social status of persons. Semantic processing of familiar faces could entail progressive stages of information abstraction. However, the cortical mechanisms supporting multistage processing of familiar faces have not been characterized. Here, using an event-related fMRI experiment, familiar faces from four celebrity groups (actors, singers, politicians, and football players) and unfamiliar faces were presented to the human subjects (both males and females) while they were engaged in a face categorization task. We systematically explored the cortical representations for faces, familiar faces, subcategories of familiar faces, and familiar face identities using whole-brain univariate analysis, searchlight-based multivariate pattern analysis (MVPA), and functional connectivity analysis. Convergent evidence from all these analyses revealed a set of overlapping regions within posterior cingulate cortex (PCC) that contained decodable fMRI responses for representing different levels of semantic knowledge about familiar faces. Our results suggest a multistage pathway in PCC for processing semantic information of faces, analogous to the multistage pathway in ventral temporal cortex for processing perceptual information of faces.SIGNIFICANCE STATEMENT Recognizing familiar faces is an important component of social communications. Previous research has shown that a distributed network of brain areas is involved in processing the semantic information of familiar faces. However, it is not clear how different levels of semantic information are represented in the brain. Here, we evaluated the multivariate response patterns across the entire cortex to discover the areas that contain information for familiar faces, subcategories of familiar faces, and identities of familiar faces. The searchlight maps revealed that different levels of semantic information are represented in topographically adjacent areas within posterior cingulate cortex (PCC). The results suggest that semantic processing of faces is mediated through progressive stages of information abstraction in PCC.

Image or Identity? Only Super-recognizers’ (Memor)Ability is Consistently Viewpoint-Invariant

A face’s memorability refers to the unique combination of visual features facilitating its recognition. Despite considerable variation in face recognition ability amongst the general population, individuals show substantial concordance regarding faces’ memorability. This agreement persists, though reduced, when the viewpoints across which identities are seen at encoding and recognition differ. Consequently, individuals must extract some invariant facial information during recognition, robust to changes in viewpoint, to do so consistently (i.e. as a function of stimulus memorability). However, the extent of such consistency remains unclear. Therefore, in two experiments we tested recognition of (i) implicitly encoded face images and (ii) explicitly encoded identities in a group of control observers against a group of “Super-Recognizers” (SRs) possessing exceptional face processing skills (Ramon, 2021). Novel “With or Without You” (WoWY) resampling analyses assessed the consistency of SRs relative to control observers, simultaneously providing measures of intraindividual consistency. When implicit encoding was surreptitiously solicited (Experiment 1), recognition of studied images was comparable between groups. Yet, when encoding was explicitly solicited (Experiment 2), SRs more accurately recognized identities across viewpoint changes than controls. Critically, image-dependent information could only inform recognition in the first experiment, whereas viewpoint-invariant information could inform recognition equally in both. Individual observers’ performance profiles reveal that only SRs memorability-related performance was consistent between experiments. Overall, SRs’ unique capacity for recognizing faces based on viewpoint-invariant information is rooted in fundamentally more robust representations of identity. These results invite a reinterpretation of face memorability describing viewpoint-invariant information, diagnostic of facial identity representations in memory. <strong>Highlights</strong> <ul><li>Super-Recognizers (SRs) do not uniformly recognize face images better than controls; both SRs and control observers tend to recognize high- and low-memorability stimuli equally well.</li><li>Novel “With or Without You” analyses, combining split-half resampling and permutation techniques, revealed inter- and intra-individual consistency in recognition memory.</li><li>SRs excelled at explicit (but not implicit) learning of facial identities relative to controls when viewpoint-invariant information was task-relevant.</li><li>SRs consistently represent faces with identity-based (viewpoint-invariant) memorability, while controls use either image- or identity-based representations.<br />The finding that consistency of identity-based memorability is relatively unique to SRs suggests that it constitutes a potential marker of overall face recognition ability.</li></ul>

Featural and configural processing of faces and houses in matched dyslexic and typical readers

While dyslexia is typically described as a phonological deficit, recent evidence suggests that ventral stream regions, important for visual categorization and object recognition, are hypoactive in dyslexic readers who might accordingly show visual recognition deficits. By manipulating featural and configural information of faces and houses, we investigated whether dyslexic readers are disadvantaged at recognizing certain object classes or using particular visual processing mechanisms. Dyslexic readers found it harder to recognize objects (houses), suggesting that visual problems in dyslexia are not completely domain-specific. Face recognition accuracy was equivalent in the two groups. Lower recognition accuracy for houses was also related to reading difficulties even when accuracy for faces was kept constant, which could indicate a specific relationship between visual word processing and visual processing of non-face objects. Representational similarity analyses (RSA) revealed that featural and configural processes were clearly separable in typical readers, which was not the case for dyslexic readers who appear to rely on a single process. This was not restricted to particular visual categories, occurring for both faces and houses. We speculate that reading deficits in some dyslexic readers reflect their reliance on a single process for object recognition.

Olfaction in the Multisensory Processing of Faces: A Narrative Review of the Influence of Human Body Odors.

A recent body of research has emerged regarding the interactions between olfaction and other sensory channels to process social information. The current review examines the influence of body odors on face perception, a core component of human social cognition. First, we review studies reporting how body odors interact with the perception of invariant facial information (i.e., identity, sex, attractiveness, trustworthiness, and dominance). Although we mainly focus on the influence of body odors based on axillary odor, we also review findings about specific steroids present in axillary sweat (i.e., androstenone, androstenol, androstadienone, and estratetraenol). We next survey the literature showing body odor influences on the perception of transient face properties, notably in discussing the role of body odors in facilitating or hindering the perception of emotional facial expression, in relation to competing frameworks of emotions. Finally, we discuss the developmental origins of these olfaction-to-vision influences, as an emerging literature indicates that odor cues strongly influence face perception in infants. Body odors with a high social relevance such as the odor emanating from the mother have a widespread influence on various aspects of face perception in infancy, including categorization of faces among other objects, face scanning behavior, or facial expression perception. We conclude by suggesting that the weight of olfaction might be especially strong in infancy, shaping social perception, especially in slow-maturing senses such as vision, and that this early tutoring function of olfaction spans all developmental stages to disambiguate a complex social environment by conveying key information for social interactions until adulthood.