Detecting cooperative partners in situations that have financial stakes is crucial to successful social exchange. The authors tested whether humans are sensitive subtle facial dynamics of counterparts when deciding trust and cooperate. Participants played a 2-person game before which the other player were manipulated using brief (<6 s) but highly realistic animations. Results showed significantly influenced participants' (a) choice with whom play (b) decisions It was also found inferences...

Abstract Human movement researchers are often restricted to laboratory environments and data capture techniques that time and/or resource intensive. Markerless pose estimation algorithms show great potential facilitate large scale studies ‘in the wild’, i.e., outside of constraints imposed by marker-based motion capture. However, accuracy such has not yet been fully evaluated. We computed 3D joint centre locations using several pre-trained deep-learning based methods (OpenPose, AlphaPose,...

Current unsupervised image-to-image translation techniques struggle to focus their attention on individual objects without altering the background or way multiple interact within a scene. Motivated by important role of in human perception, we tackle this limitation introducing mechanisms that are jointly adversarialy trained with generators and discriminators. We demonstrate qualitatively quantitatively our approach is able attend relevant regions image requiring supervision, doing so it...

This paper presents the first dynamic 3D FACS data set for facial expression research, containing 10 subjects performing between 19 and 97 different AUs both individually in combination. In total corpus contains 519 AU sequences. The peak frame of each sequence has been manually coded by certified experts. provides a ground truth based recognition systems. order to use this data, we describe framework building morphable models. includes novel Active Appearance Model (AAM) registration mesh...

We present an interactive, robust and high quality method for fast shadow removal.To perform detection we use on-the-fly learning approach guided by two rough user inputs the pixels of lit area.From this derive a fusion image that magnifies boundary intensity change due to illumination variation.After detection, removal registering penumbra normalised frame which allows us efficiently estimate non-uniform changes, resulting in accurate removal.We also first reliable, validated multi-scene...

Speech-driven 3D facial animation has been widely explored, with applications in gaming, character animation, virtual reality, and telepresence systems. State-of-the-art methods deform the face topology of target actor to sync input audio without considering identity-specific speaking style idiosyncrasies, thus, resulting unrealistic inaccurate lip movements. To address this, we present Imitator, a speech-driven expression synthesis method, which learns details from short video produces...

Perceptually lossless foveated rendering methods exploit human perception by selectively at different quality levels based on eye gaze (at a lower computational cost) while still maintaining the user's of full render. We consider three and propose practical rules thumb for each method to achieve significant performance gains in real-time frameworks. Additionally, we contribute new metric perceptual building HDR-VDP2 that, unlike traditional metrics, considers loss fidelity peripheral vision...

The automatic extraction of animal 3D pose from images without markers is interest in a range scientific fields. Most work to date predicts RGB images, based on 2D labelling joint positions. However, due the difficult nature obtaining training data, no ground truth dataset motion available quantitatively evaluate these approaches. In addition, lack data also makes it train pose-prediction methods similar manner popular field body-pose prediction. our work, we focus problem canine estimation...

The increased availability and maturity of head-mounted wearable devices opens up opportunities for remote communication collaboration. However, the signal streams provided by these (e.g., head pose, hand gaze direction) do not represent a whole person. One main open problems is therefore how to leverage signals build faithful representations user. In this paper, we propose method based on variational autoencoders generate articulated poses human skeleton noisy pose. Our approach relies...

This study presented a fully automated deep learning based markerless motion capture workflow and evaluated its performance against marker-based during overground running, walking counter movement jumping. Multi-view high speed (200 Hz) image data were collected concurrently with (criterion data), permitting direct comparison between methods. Lower limb kinematic for 15 participants computed using 2D pose estimation, our 3D fusion process OpenSim inverse kinematics modelling. Results...

Embodying virtual twins – photorealistic and personalized avatars will soon be easily achievable in consumer-grade VR. For the first time, we explored how photorealism personalization impact self-identification, as well embodiment, avatar perception presence. Twenty participants were individually scanned and, a two-hour session, embodied four (high personalized, low high generic, generic). Questionnaire responses revealed stronger mid-immersion body ownership for compared to all other types,...

A user-centric method for fast, interactive, robust and high-quality shadow removal is presented. Our algorithm can perform detection in a range of difficult cases: such as highly textured colored shadows. To an on-the-fly learning approach adopted guided by two rough user inputs the pixels lit area. After detection, performed registering penumbra to normalized frame which allows us efficient estimation non-uniform illumination changes, resulting accurate removal. Another major contribution...

The ability to accurately and non-invasively measure 3D mass centre positions their derivatives can provide rich insight into the physical demands of sports training competition. This study examines a method for measuring velocities using markerless human pose estimation Kalman smoothing. Marker (Qualysis) (OpenPose) motion capture data were captured synchronously sprinting skeleton push starts. Mass derived from raw contained large errors both pushing (mean ± SD = 0.127 0.943 −0.197 1.549...

We investigate the problem of integrating complementary audio and visual modalities for speech separation. Rather than using independence criteria suggested in most blind source separation (BSS) systems, we use features from a video signal as additional information to optimize unmixing matrix. achieve this by statistical model characterizing nonlinear coherence between criterion both instantaneous convolutive mixtures. acquire applying Bayesian framework fused feature observations based on...

Although the human face is commonly used as a physiological biometric, very little work has been done to exploit idiosyncrasies of facial motions for person identification. In this paper, we investigate <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">uniqueness</i> and xmlns:xlink="http://www.w3.org/1999/xlink">permanence</i> actions determine whether these can be behavioral biometric. Experiments are carried out using 3-D video data participants...

We introduce a video-based approach for producing water surface models. Recent advances in this field output high-quality results but require dedicated capturing devices and only work limited conditions. In contrast, our method achieves good tradeoff between the visual quality production cost: It automatically produces visually plausible animation using single viewpoint video as input. Our is based on two discoveries: first, shape from shading (SFS) adequate to capture appearance dynamic...

Automatic perception of human behaviors during social interactions is crucial for AR/VR applications, and an essential component estimation plausible 3D pose shape our partners from the egocentric view. One biggest challenges this task severe body truncation due to close distances in scenarios, which brings large ambiguities unseen parts. To tackle challenge, we propose a novel scene-conditioned diffusion method model distribution. Conditioned on scene geometry, generates bodies humanscene...

To objectively quantify facial movement in response to expression and spoken word.Experimental study. Setting - Department of Dental Health Biological Sciences, University Hospital, Cardiff, UK.Facial was assessed a standardized smile the utterance 'puppy'. The sequences were recorded using non-invasive, three-dimensional motion analysis image capture system (3dMDface Dynamic System) at 48 frames per second.To movement, sequential sequence aligned baseline/reference frame three-dimensionally...

This study examined if occluded joint locations, obtained from 2D markerless motion capture (single camera view), produced angles with reduced agreement compared to visible joints, and frontal plane were usable for practical applications. Fifteen healthy participants performed over-ground walking whilst recorded by fifteen marker-based cameras two machine vision (frontal sagittal plane). Repeated measures Bland-Altman analysis illustrated that standard deviation of bias limits the...

Generating both plausible and accurate full body avatar motion is the key to quality of immersive experiences in mixed reality scenarios. Head-Mounted Devices (HMDs) typically only provide a few input signals, such as head hands 6-DoF. Recently, different approaches achieved impressive performance generating given signal. However, best our knowledge, all existing rely on hand visibility. While this case when, e.g., using controllers, considerable proportion do not involve controllers instead...

In this paper we present the first Facial Action Coding System (FACS) valid model to be based on dynamic 3D scans of human faces for use in graphics and psychological research. The consists FACS Unit (AU) parameters has been independently validated by experts. Using model, explore perceptual differences between linear facial motions -- represented a blend shape approach real that have synthesized through model. Through numerical measures visualizations, show latter type motion is...