Chirag Raman

Erroll Wood, Tadas Baltrusaitis, Charlie Hewitt et al.

Landmarks often play a key role in face analysis, but many aspects of identity or expression cannot be represented by sparse landmarks alone. Thus, in order to reconstruct faces more accurately, landmarks are often combined with additional signals like depth images or techniques like differentiable rendering. Can we keep things simple by just using more landmarks? In answer, we present the first method that accurately predicts 10x as many landmarks as usual, covering the whole head, including the eyes and teeth. This is accomplished using synthetic training data, which guarantees perfect landmark annotations. By fitting a morphable model to these dense landmarks, we achieve state-of-the-art results for monocular 3D face reconstruction in the wild. We show that dense landmarks are an ideal signal for integrating face shape information across frames by demonstrating accurate and expressive facial performance capture in both monocular and multi-view scenarios. This approach is also highly efficient: we can predict dense landmarks and fit our 3D face model at over 150FPS on a single CPU thread. Please see our website: https://microsoft.github.io/DenseLandmarks/.

Chirag Raman, Jose Vargas-Quiros, Stephanie Tan et al.

Recording the dynamics of unscripted human interactions in the wild is challenging due to the delicate trade-offs between several factors: participant privacy, ecological validity, data fidelity, and logistical overheads. To address these, following a 'datasets for the community by the community' ethos, we propose the Conference Living Lab (ConfLab): a new concept for multimodal multisensor data collection of in-the-wild free-standing social conversations. For the first instantiation of ConfLab described here, we organized a real-life professional networking event at a major international conference. Involving 48 conference attendees, the dataset captures a diverse mix of status, acquaintance, and networking motivations. Our capture setup improves upon the data fidelity of prior in-the-wild datasets while retaining privacy sensitivity: 8 videos (1920x1080, 60 fps) from a non-invasive overhead view, and custom wearable sensors with onboard recording of body motion (full 9-axis IMU), privacy-preserving low-frequency audio (1250 Hz), and Bluetooth-based proximity. Additionally, we developed custom solutions for distributed hardware synchronization at acquisition and time-efficient continuous annotation of body keypoints and actions at high sampling rates. Our benchmarks showcase some of the open research tasks related to in-the-wild privacy-preserving social data analysis: keypoints detection from overhead camera views, skeleton-based no-audio speaker detection, and F-formation detection.

Chirag Raman, Charlie Hewitt, Erroll Wood et al.

Recent advances in synthesizing realistic faces have shown that synthetic training data can replace real data for various face-related computer vision tasks. A question arises: how important is realism? Is the pursuit of photorealism excessive? In this work, we show otherwise. We boost the realism of our synthetic faces by introducing dynamic skin wrinkles in response to facial expressions and observe significant performance improvements in downstream computer vision tasks. Previous approaches for producing such wrinkles either required prohibitive artist effort to scale across identities and expressions or were not capable of reconstructing high-frequency skin details with sufficient fidelity. Our key contribution is an approach that produces realistic wrinkles across a large and diverse population of digital humans. Concretely, we formalize the concept of mesh-tension and use it to aggregate possible wrinkles from high-quality expression scans into albedo and displacement texture maps. At synthesis, we use these maps to produce wrinkles even for expressions not represented in the source scans. Additionally, to provide a more nuanced indicator of model performance under deformations resulting from compressed expressions, we introduce the 300W-winks evaluation subset and the Pexels dataset of closed eyes and winks.

Chirag Raman, Hayley Hung, Marco Loog

Forecasting tasks surrounding the dynamics of low-level human behavior are of significance to multiple research domains. In such settings, methods for explaining specific forecasts can enable domain experts to gain insights into the predictive relationships between behaviors. In this work, we introduce and address the following question: given a probabilistic forecasting model how can we identify observed windows that the model considers salient when making its forecasts? We build upon a general definition of information-theoretic saliency grounded in human perception and extend it to forecasting settings by leveraging a crucial attribute of the domain: a single observation can result in multiple valid futures. We propose to express the saliency of an observed window in terms of the differential entropy of the resulting predicted future distribution. In contrast to existing methods that either require explicit training of the saliency mechanism or access to the internal states of the forecasting model, we obtain a closed-form solution for the saliency map for commonly used density functions in probabilistic forecasting. We empirically demonstrate how our framework can recover salient observed windows from head pose features for the sample task of speaking-turn forecasting using a synthesized conversation dataset.

Joris Kirchner, Amogh Gudi, Marian Bittner et al.

Deep learning vision models excel with abundant supervision, but many applications face label scarcity and class imbalance. Controllable image editing can augment scarce labeled data, yet edits often introduce artifacts and entangle non-target attributes. We study this in facial expression analysis, targeting Action Unit (AU) manipulation where annotation is costly and AU co-activation drives entanglement. We present a facial manipulation method that operates in the semantic latent space of a pre-trained face generator (Diffusion Autoencoder). Using lightweight linear models, we reduce entanglement of semantic features via (i) dependency-aware conditioning that accounts for AU co-activation, and (ii) orthogonal projection that removes nuisance attribute directions (e.g., glasses), together with an expression neutralization step to enable absolute AU edit. We use these edits to balance AU occurrence by editing labeled faces and to diversify identities/demographics via controlled synthesis. Augmenting AU detector training with the generated data improves accuracy and yields more disentangled predictions with fewer co-activation shortcuts, outperforming alternative data-efficient training strategies and suggesting improvements similar to what would require substantially more labeled data in our learning-curve analysis. Compared to prior methods, our edits are stronger, produce fewer artifacts, and preserve identity better.

Xiangwei Shi, Era Dorta Perez, Ruud de Jong et al.

Multi-view capture systems have been an important tool in research for recording human motion under controlling conditions. Most existing systems are specified around video streams and provide little or no support for audio acquisition and rigorous audio-video alignment, despite both being essential for studying conversational interaction where timing at the level of turn-taking, overlap, and prosody matters. In this technical report, we describe an audio-visual multi-view capture system that addresses this gap by treating synchronized audio and synchronized video as first-class signals. The system combines a multi-camera pipeline with multi-channel microphone recording under a unified timing architecture and provides a practical workflow for calibration, acquisition, and quality control that supports repeatable recordings at scale. We quantify synchronization performance in deployment and show that the resulting recordings are temporally consistent enough to support fine-grained analysis and data-driven modeling of conversation behavior.

Ojas Shirekar, Wim Pouw, Chenxu Hao et al.

Digital humans are emerging as autonomous agents in multiparty interactions, yet existing evaluation metrics largely ignore contextual coordination dynamics. We introduce a unified, intervention-driven framework for objective assessment of multiparty social behaviour in skeletal motion data, spanning three complementary dimensions: (1) synchrony via Cross-Recurrence Quantification Analysis, (2) temporal alignment via Multiscale Empirical Mode Decompositionbased Beat Consistency, and (3) structural similarity via Soft Dynamic Time Warping. We validate metric sensitivity through three theory-driven perturbations -- gesture kinematic dampening, uniform speech-gesture delays, and prosodic pitch-variance reduction-applied to $\approx 145$ 30-second thin slices of group interactions from the DnD dataset. Mixed-effects analyses reveal predictable, joint-independent shifts: dampening increases CRQA determinism and reduces beat consistency, delays weaken cross-participant coupling, and pitch flattening elevates F0 Soft-DTW costs. A complementary perception study ($N=27$) compares judgments of full-video and skeleton-only renderings to quantify representation effects. Our three measures deliver orthogonal insights into spatial structure, timing alignment, and behavioural variability. Thereby forming a robust toolkit for evaluating and refining socially intelligent agents. Code available on \href{https://github.com/tapri-lab/gig-interveners}{GitHub}.

Felix Dollack, Kiyoshi Kiyokawa, Huakun Liu et al.

The congruence between affective experiences and physiological changes has been a debated topic for centuries. Recent technological advances in measurement and data analysis provide hope to solve this epic challenge. Open science and open data practices, together with data analysis challenges open to the academic community, are also promising tools for solving this problem. In this entry to the Emotion Physiology and Experience Collaboration (EPiC) challenge, we propose a data analysis solution that combines theoretical assumptions with data-driven methodologies. We used feature engineering and ensemble selection. Each predictor was trained on subsets of the training data that would maximize the information available for training. Late fusion was used with an averaging step. We chose to average considering a ``wisdom of crowds'' strategy. This strategy yielded an overall RMSE of 1.19 in the test set. Future work should carefully explore if our assumptions are correct and the potential of weighted fusion.

Jose Vargas Quiros, Chirag Raman, Stephanie Tan et al.

Recognizing speaking in humans is a central task towards understanding social interactions. Ideally, speaking would be detected from individual voice recordings, as done previously for meeting scenarios. However, individual voice recordings are hard to obtain in the wild, especially in crowded mingling scenarios due to cost, logistics, and privacy concerns. As an alternative, machine learning models trained on video and wearable sensor data make it possible to recognize speech by detecting its related gestures in an unobtrusive, privacy-preserving way. These models themselves should ideally be trained using labels obtained from the speech signal. However, existing mingling datasets do not contain high quality audio recordings. Instead, speaking status annotations have often been inferred by human annotators from video, without validation of this approach against audio-based ground truth. In this paper we revisit no-audio speaking status estimation by presenting the first publicly available multimodal dataset with high-quality individual speech recordings of 33 subjects in a professional networking event. We present three baselines for no-audio speaking status segmentation: a) from video, b) from body acceleration (chest-worn accelerometer), c) from body pose tracks. In all cases we predict a 20Hz binary speaking status signal extracted from the audio, a time resolution not available in previous datasets. In addition to providing the signals and ground truth necessary to evaluate a wide range of speaking status detection methods, the availability of audio in REWIND makes it suitable for cross-modality studies not feasible with previous mingling datasets. Finally, our flexible data consent setup creates new challenges for multimodal systems under missing modalities.

Prerana Achar, Mayank Patel, Anushka Mulik et al.

We present a comparison review that evaluates popular techniques for garment draping for 3D fashion design, virtual try-ons, and animations. A comparative study is performed between various methods for garment draping of clothing over the human body. These include numerous models, such as physics and machine learning based techniques, collision handling, and more. Performance evaluations and trade-offs are discussed to ensure informed decision-making when choosing the most appropriate approach. These methods aim to accurately represent deformations and fine wrinkles of digital garments, considering the factors of data requirements, and efficiency, to produce realistic results. The research can be insightful to researchers, designers, and developers in visualizing dynamic multi-layered 3D clothing.

Bernd Dudzik, Tiffany Matej Hrkalovic, Chenxu Hao et al.

Automatic Affect Prediction (AAP) uses computational analysis of input data such as text, speech, images, and physiological signals to predict various affective phenomena (e.g., emotions or moods). These models are typically constructed using supervised machine-learning algorithms, which rely heavily on labeled training datasets. In this position paper, we posit that all AAP training data are derived from human Affective Interpretation Processes, resulting in a form of Affective Meaning. Research on human affect indicates a form of complexity that is fundamental to such meaning: it can possess what we refer to here broadly as Qualities of Indeterminacy (QIs) - encompassing Subjectivity (meaning depends on who is interpreting), Uncertainty (lack of confidence regarding meanings' correctness), Ambiguity (meaning contains mutually exclusive concepts) and Vagueness (meaning is situated at different levels in a nested hierarchy). Failing to appropriately consider QIs leads to results incapable of meaningful and reliable predictions. Based on this premise, we argue that a crucial step in adequately addressing indeterminacy in AAP is the development of data collection practices for modeling corpora that involve the systematic consideration of 1) a relevant set of QIs and 2) context for the associated interpretation processes. To this end, we are 1) outlining a conceptual model of AIPs and the QIs associated with the meaning these produce and a conceptual structure of relevant context, supporting understanding of its role. Finally, we use our framework for 2) discussing examples of context-sensitivity-related challenges for addressing QIs in data collection setups. We believe our efforts can stimulate a structured discussion of both the role of aspects of indeterminacy and context in research on AAP, informing the development of better practices for data collection and analysis.

Augustinas Jučas, Chirag Raman

Adaptively forecasting human behavior in social settings is an important step toward achieving Artificial General Intelligence. Most existing research in social forecasting has focused either on unfocused interactions, such as pedestrian trajectory prediction, or on monadic and dyadic behavior forecasting. In contrast, social psychology emphasizes the importance of group interactions for understanding complex social dynamics. This creates a gap that we address in this paper: forecasting social interactions at the group (conversation) level. Additionally, it is important for a forecasting model to be able to adapt to groups unseen at train time, as even the same individual behaves differently across different groups. This highlights the need for a forecasting model to explicitly account for each group's unique dynamics. To achieve this, we adopt a meta-learning approach to human behavior forecasting, treating every group as a separate meta-learning task. As a result, our method conditions its predictions on the specific behaviors within the group, leading to generalization to unseen groups. Specifically, we introduce Social Process (SP) models, which predict a distribution over future multimodal cues jointly for all group members based on their preceding low-level multimodal cues, while incorporating other past sequences of the same group's interactions. In this work we also analyze the generalization capabilities of SP models in both their outputs and latent spaces through the use of realistic synthetic datasets.

Mircea Lică, Ojas Shirekar, Baptiste Colle et al.

Embodied agents powered by large language models (LLMs), such as Voyager, promise open-ended competence in worlds such as Minecraft. However, when powered by open-weight LLMs they still falter on elementary tasks after domain-specific fine-tuning. We propose MindForge, a generative-agent framework for cultural lifelong learning through explicit perspective taking. We introduce three key innovations: (1) a structured theory of mind representation linking percepts, beliefs, desires, and actions; (2) natural inter-agent communication; and (3) a multi-component memory system. Following the cultural learning framework, we test MindForge in both instructive and collaborative settings within Minecraft. In an instructive setting with GPT-4, MindForge agents powered by open-weight LLMs significantly outperform their Voyager counterparts in basic tasks yielding $3\times$ more tech-tree milestones and collecting $2.3\times$ more unique items than the Voyager baseline. Furthermore, in fully \textit{collaborative} settings, we find that the performance of two underachieving agents improves with more communication rounds, echoing the Condorcet Jury Theorem. MindForge agents demonstrate sophisticated behaviors, including expert-novice knowledge transfer, collaborative problem solving, and adaptation to out-of-distribution tasks through accumulated cultural experiences.

Maria Tsfasman, Avinash Saravanan, Dekel Viner et al.

How human-like do conversational robots need to look to enable long-term human-robot conversation? One essential aspect of long-term interaction is a human's ability to adapt to the varying degrees of a conversational partner's engagement and emotions. Prosodically, this can be achieved through (dis)entrainment. While speech-synthesis has been a limiting factor for many years, restrictions in this regard are increasingly mitigated. These advancements now emphasise the importance of studying the effect of robot embodiment on human entrainment. In this study, we conducted a between-subjects online human-robot interaction experiment in an educational use-case scenario where a tutor was either embodied through a human or a robot face. 43 English-speaking participants took part in the study for whom we analysed the degree of acoustic-prosodic entrainment to the human or robot face, respectively. We found that the degree of subjective and objective perception of anthropomorphism positively correlates with acoustic-prosodic entrainment.

Chirag Raman, Hayley Hung, Marco Loog

Free-standing social conversations constitute a yet underexplored setting for human behavior forecasting. While the task of predicting pedestrian trajectories has received much recent attention, an intrinsic difference between these settings is how groups form and disband. Evidence from social psychology suggests that group members in a conversation explicitly self-organize to sustain the interaction by adapting to one another's behaviors. Crucially, the same individual is unlikely to adapt similarly across different groups; contextual factors such as perceived relationships, attraction, rapport, etc., influence the entire spectrum of participants' behaviors. A question arises: how can we jointly forecast the mutually dependent futures of conversation partners by modeling the dynamics unique to every group? In this paper, we propose the Social Process (SP) models, taking a novel meta-learning and stochastic perspective of group dynamics. Training group-specific forecasting models hinders generalization to unseen groups and is challenging given limited conversation data. In contrast, our SP models treat interaction sequences from a single group as a meta-dataset: we condition forecasts for a sequence from a given group on other observed-future sequence pairs from the same group. In this way, an SP model learns to adapt its forecasts to the unique dynamics of the interacting partners, generalizing to unseen groups in a data-efficient manner. Additionally, we first rethink the task formulation itself, motivating task requirements from social science literature that prior formulations have overlooked. For our formulation of Social Cue Forecasting, we evaluate the empirical performance of our SP models against both non-meta-learning and meta-learning approaches with similar assumptions. The SP models yield improved performance on synthetic and real-world behavior datasets.

Navin Raj Prabhu, Chirag Raman, Hayley Hung

Social interactions in general are multifaceted and there exists a wide set of factors and events that influence them. In this paper, we quantify social interactions with a holistic viewpoint on individual experiences, particularly focusing on non-task-directed spontaneous interactions. To achieve this, we design a novel perceived measure, the perceived Conversation Quality, which intends to quantify spontaneous interactions by accounting for several socio-dimensional aspects of individual experiences. To further quantitatively study spontaneous interactions, we devise a questionnaire which measures the perceived Conversation Quality, at both the individual- and at the group- level. Using the questionnaire, we collected perceived annotations for conversation quality in a publicly available dataset using naive annotators. The results of the analysis performed on the distribution and the inter-annotator agreeability shows that naive annotators tend to agree less in cases of low conversation quality samples, especially while annotating for group-level conversation quality.

Chirag Raman, Stephanie Tan, Hayley Hung

Existing data acquisition literature for human behavior research provides wired solutions, mainly for controlled laboratory setups. In uncontrolled free-standing conversation settings, where participants are free to walk around, these solutions are unsuitable. While wireless solutions are employed in the broadcasting industry, they can be prohibitively expensive. In this work, we propose a modular and cost-effective wireless approach for synchronized multisensor data acquisition of social human behavior. Our core idea involves a cost-accuracy trade-off by using Network Time Protocol (NTP) as a source reference for all sensors. While commonly used as a reference in ubiquitous computing, NTP is widely considered to be insufficiently accurate as a reference for video applications, where Precision Time Protocol (PTP) or Global Positioning System (GPS) based references are preferred. We argue and show, however, that the latency introduced by using NTP as a source reference is adequate for human behavior research, and the subsequent cost and modularity benefits are a desirable trade-off for applications in this domain. We also describe one instantiation of the approach deployed in a real-world experiment to demonstrate the practicality of our setup in-the-wild.

Chirag Raman, Hayley Hung

The detection of free-standing conversing groups has received significant attention in recent years. In the absence of a formal definition, most studies operationalize the notion of a conversation group either through a spatial or a temporal lens. Spatially, the most commonly used representation is the F-formation, defined by social scientists as the configuration in which people arrange themselves to sustain an interaction. However, the use of this representation is often accompanied with the simplifying assumption that a single conversation occurs within an F-formation. Temporally, various categories have been used to organize conversational units; these include, among others, turn, topic, and floor. Some of these concepts are hard to define objectively by themselves. The present work constitutes an initial exploration into unifying these perspectives by primarily posing the question: can we use the observation of simultaneous speaker turns to infer whether multiple conversation floors exist within an F-formation? We motivate a metric for the existence of distinct conversation floors based on simultaneous speaker turns, and provide an analysis using this metric to characterize conversations across F-formations of varying cardinality. We contribute two key findings: firstly, at the average speaking turn duration of about two seconds for humans, there is evidence for the existence of multiple floors within an F-formation; and secondly, an increase in the cardinality of an F-formation correlates with a decrease in duration of simultaneous speaking turns.

Yulun Du, Chirag Raman, Alan W Black et al.

Distracted driving is deadly, claiming 3,477 lives in the U.S. in 2015 alone. Although there has been a considerable amount of research on modeling the distracted behavior of drivers under various conditions, accurate automatic detection using multiple modalities and especially the contribution of using the speech modality to improve accuracy has received little attention. This paper introduces a new multimodal dataset for distracted driving behavior and discusses automatic distraction detection using features from three modalities: facial expression, speech and car signals. Detailed multimodal feature analysis shows that adding more modalities monotonically increases the predictive accuracy of the model. Finally, a simple and effective multimodal fusion technique using a polynomial fusion layer shows superior distraction detection results compared to the baseline SVM and neural network models.