Sangwu Lee

Md Saiful Islam, Wasifur Rahman, Abdelrahman Abdelkader et al.

We present an artificial intelligence system to remotely assess the motor performance of individuals with Parkinson's disease (PD). Participants performed a motor task (i.e., tapping fingers) in front of a webcam, and data from 250 global participants were rated by three expert neurologists following the Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS). The neurologists' ratings were highly reliable, with an intra-class correlation coefficient (ICC) of 0.88. We developed computer algorithms to obtain objective measurements that align with the MDS-UPDRS guideline and are strongly correlated with the neurologists' ratings. Our machine learning model trained on these measures outperformed an MDS-UPDRS certified rater, with a mean absolute error (MAE) of 0.59 compared to the rater's MAE of 0.79. However, the model performed slightly worse than the expert neurologists (0.53 MAE). The methodology can be replicated for similar motor tasks, providing the possibility of evaluating individuals with PD and other movement disorders remotely, objectively, and in areas with limited access to neurological care.

Tariq Adnan, Md Saiful Islam, Wasifur Rahman et al.

We present an efficient and accessible PD screening method by leveraging AI-driven models enabled by the largest video dataset of facial expressions from 1,059 unique participants. This dataset includes 256 individuals with PD, 165 clinically diagnosed, and 91 self-reported. Participants used webcams to record themselves mimicking three facial expressions (smile, disgust, and surprise) from diverse sources encompassing their homes across multiple countries, a US clinic, and a PD wellness center in the US. Facial landmarks are automatically tracked from the recordings to extract features related to hypomimia, a prominent PD symptom characterized by reduced facial expressions. Machine learning algorithms are trained on these features to distinguish between individuals with and without PD. The model was tested for generalizability on external (unseen during training) test videos collected from a US clinic and Bangladesh. An ensemble of machine learning models trained on smile videos achieved an accuracy of 87.9+-0.1% (95% Confidence Interval) with an AUROC of 89.3+-0.3% as evaluated on held-out data (using k-fold cross-validation). In external test settings, the ensemble model achieved 79.8+-0.6% accuracy with 81.9+-0.3% AUROC on the clinical test set and 84.9+-0.4% accuracy with 81.2+-0.6% AUROC on participants from Bangladesh. In every setting, the model was free from detectable bias across sex and ethnic subgroups, except in the cohorts from Bangladesh, where the model performed significantly better for female participants than males. Smiling videos can effectively differentiate between individuals with and without PD, offering a potentially easy, accessible, and cost-efficient way to screen for PD, especially when a clinical diagnosis is difficult to access.

Md Kamrul Hasan, Md Saiful Islam, Sangwu Lee et al.

Pre-trained large language models have recently achieved ground-breaking performance in a wide variety of language understanding tasks. However, the same model can not be applied to multimodal behavior understanding tasks (e.g., video sentiment/humor detection) unless non-verbal features (e.g., acoustic and visual) can be integrated with language. Jointly modeling multiple modalities significantly increases the model complexity, and makes the training process data-hungry. While an enormous amount of text data is available via the web, collecting large-scale multimodal behavioral video datasets is extremely expensive, both in terms of time and money. In this paper, we investigate whether large language models alone can successfully incorporate non-verbal information when they are presented in textual form. We present a way to convert the acoustic and visual information into corresponding textual descriptions and concatenate them with the spoken text. We feed this augmented input to a pre-trained BERT model and fine-tune it on three downstream multimodal tasks: sentiment, humor, and sarcasm detection. Our approach, TextMI, significantly reduces model complexity, adds interpretability to the model's decision, and can be applied for a diverse set of tasks while achieving superior (multimodal sarcasm detection) or near SOTA (multimodal sentiment analysis and multimodal humor detection) performance. We propose TextMI as a general, competitive baseline for multimodal behavioral analysis tasks, particularly in a low-resource setting.

Md Saiful Islam, Sangwu Lee, Abdelrahman Abdelkader et al.

We present a web-based framework to screen for Parkinson's disease (PD) by allowing users to perform neurological tests in their homes. Our web framework guides the users to complete three tasks involving speech, facial expression, and finger movements. The task videos are analyzed to classify whether the users show signs of PD. We present the results in an easy-to-understand manner, along with personalized resources to further access to treatment and care. Our framework is accessible by any major web browser, improving global access to neurological care.

Syeda Nahida Akter, Sangwu Lee, Yingshan Chang et al. · cmu

Verifying a question's validity before answering is crucial in real-world applications, where users may provide imperfect instructions. In this scenario, an ideal model should address the discrepancies in the query and convey them to the users rather than generating the best possible answer. Addressing this requirement, we introduce a new compositional visual question-answering dataset, VISREAS, that consists of answerable and unanswerable visual queries formulated by traversing and perturbing commonalities and differences among objects, attributes, and relations. VISREAS contains 2.07M semantically diverse queries generated automatically using Visual Genome scene graphs. The unique feature of this task, validating question answerability with respect to an image before answering, and the poor performance of state-of-the-art models inspired the design of a new modular baseline, LOGIC2VISION that reasons by producing and executing pseudocode without any external modules to generate the answer. LOGIC2VISION outperforms generative models in VISREAS (+4.82% over LLaVA-1.5; +12.23% over InstructBLIP) and achieves a significant gain in performance against the classification models.

Donghoon Ahn, Jiwon Kang, Sanghyun Lee et al.

Recent guidance methods in diffusion models steer reverse sampling by perturbing the model to construct an implicit weak model and guide generation away from it. Among these approaches, attention perturbation has demonstrated strong empirical performance in unconditional scenarios where classifier-free guidance is not applicable. However, existing attention perturbation methods lack principled approaches for determining where perturbations should be applied, particularly in Diffusion Transformer (DiT) architectures where quality-relevant computations are distributed across layers. In this paper, we investigate the granularity of attention perturbations, ranging from the layer level down to individual attention heads, and discover that specific heads govern distinct visual concepts such as structure, style, and texture quality. Building on this insight, we propose "HeadHunter", a systematic framework for iteratively selecting attention heads that align with user-centric objectives, enabling fine-grained control over generation quality and visual attributes. In addition, we introduce SoftPAG, which linearly interpolates each selected head's attention map toward an identity matrix, providing a continuous knob to tune perturbation strength and suppress artifacts. Our approach not only mitigates the oversmoothing issues of existing layer-level perturbation but also enables targeted manipulation of specific visual styles through compositional head selection. We validate our method on modern large-scale DiT-based text-to-image models including Stable Diffusion 3 and FLUX.1, demonstrating superior performance in both general quality enhancement and style-specific guidance. Our work provides the first head-level analysis of attention perturbation in diffusion models, uncovering interpretable specialization within attention layers and enabling practical design of effective perturbation strategies.

Md Saiful Islam, Tariq Adnan, Jan Freyberg et al.

Limited accessibility to neurological care leads to underdiagnosed Parkinson's Disease (PD), preventing early intervention. Existing AI-based PD detection methods primarily focus on unimodal analysis of motor or speech tasks, overlooking the multifaceted nature of the disease. To address this, we introduce a large-scale, multi-task video dataset consisting of 1102 sessions (each containing videos of finger tapping, facial expression, and speech tasks captured via webcam) from 845 participants (272 with PD). We propose a novel Uncertainty-calibrated Fusion Network (UFNet) that leverages this multimodal data to enhance diagnostic accuracy. UFNet employs independent task-specific networks, trained with Monte Carlo Dropout for uncertainty quantification, followed by self-attended fusion of features, with attention weights dynamically adjusted based on task-specific uncertainties. To ensure patient-centered evaluation, the participants were randomly split into three sets: 60% for training, 20% for model selection, and 20% for final performance evaluation. UFNet significantly outperformed single-task models in terms of accuracy, area under the ROC curve (AUROC), and sensitivity while maintaining non-inferior specificity. Withholding uncertain predictions further boosted the performance, achieving 88.0+-0.3%$ accuracy, 93.0+-0.2% AUROC, 79.3+-0.9% sensitivity, and 92.6+-0.3% specificity, at the expense of not being able to predict for 2.3+-0.3% data (+- denotes 95% confidence interval). Further analysis suggests that the trained model does not exhibit any detectable bias across sex and ethnic subgroups and is most effective for individuals aged between 50 and 80. Requiring only a webcam and microphone, our approach facilitates accessible home-based PD screening, especially in regions with limited healthcare resources.

Masum Hasan, Cengiz Ozel, Nina Long et al.

Hand pose estimation plays a vital role in capturing subtle nonverbal cues essential for understanding human affect. However, collecting diverse, expressive real-world data remains challenging due to labor-intensive manual annotation that often underrepresents demographic diversity and natural expressions. To address this issue, we introduce a cost-effective approach to generating synthetic data using high-fidelity 3D hand models and a wide range of affective hand poses. Our method includes varied skin tones, genders, dynamic environments, realistic lighting conditions, and diverse naturally occurring gesture animations. The resulting dataset, Hi5, contains 583,000 pose-annotated images, carefully balanced to reflect natural diversity and emotional expressiveness. Models trained exclusively on Hi5 achieve performance comparable to human-annotated datasets, exhibiting superior robustness to occlusions and consistent accuracy across diverse skin tones -- which is crucial for reliably recognizing expressive gestures in affective computing applications. Our results demonstrate that synthetic data effectively addresses critical limitations of existing datasets, enabling more inclusive, expressive, and reliable gesture recognition systems while achieving competitive performance in pose estimation benchmarks. The Hi5 dataset, data synthesis pipeline, source code, and game engine project are publicly released to support further research in synthetic hand-gesture applications.

Syeda Nahida Akter, Aman Madaan, Sangwu Lee et al.

The potential of Vision-Language Models (VLMs) often remains underutilized in handling complex text-based problems, particularly when these problems could benefit from visual representation. Resonating with humans' ability to solve complex text-based problems by (1) creating a visual diagram from the problem and (2) deducing what steps they need to take to solve it, we propose Self-Imagine. We leverage a single Vision-Language Model (VLM) to generate a structured representation of the question using HTML, then render the HTML as an image, and finally use the same VLM to answer the question using both the question and the image. Our approach does not require any additional training data or training. We evaluate our approach on three mathematics tasks and nine general-purpose reasoning tasks using state-of-the-art (LLAVA-1.5 and GEMINI PRO) VLMs. Our approach boosts the performance of LLAVA-1.5 and GEMINI PRO on all math tasks (on average GSM8K: +3.1%; ASDIV: +3.2%; SVAMP: +6.9%) and the majority of the general-purpose reasoning tasks by 3.2% to 6.0% on average.

Wasifur Rahman, Sangwu Lee, Md. Saiful Islam et al.

In this paper, we envision a web-based framework that can help anyone, anywhere around the world record a short speech task, and analyze the recorded data to screen for Parkinson's disease (PD). We collected data from 726 unique participants (262 PD, 38% female; 464 non-PD, 65% female; average age: 61) -- from all over the US and beyond. A small portion of the data was collected in a lab setting to compare quality. The participants were instructed to utter a popular pangram containing all the letters in the English alphabet "the quick brown fox jumps over the lazy dog..". We extracted both standard acoustic features (Mel Frequency Cepstral Coefficients (MFCC), jitter and shimmer variants) and deep learning based features from the speech data. Using these features, we trained several machine learning algorithms. We achieved 0.75 AUC (Area Under The Curve) performance on determining presence of self-reported Parkinson's disease by modeling the standard acoustic features through the XGBoost -- a gradient-boosted decision tree model. Further analysis reveal that the widely used MFCC features and a subset of previously validated dysphonia features designed for detecting Parkinson's from verbal phonation task (pronouncing 'ahh') contains the most distinct information. Our model performed equally well on data collected in controlled lab environment as well as 'in the wild' across different gender and age groups. Using this tool, we can collect data from almost anyone anywhere with a video/audio enabled device, contributing to equity and access in neurological care.

Wasifur Rahman, Md. Kamrul Hasan, Sangwu Lee et al.

Recent Transformer-based contextual word representations, including BERT and XLNet, have shown state-of-the-art performance in multiple disciplines within NLP. Fine-tuning the trained contextual models on task-specific datasets has been the key to achieving superior performance downstream. While fine-tuning these pre-trained models is straightforward for lexical applications (applications with only language modality), it is not trivial for multimodal language (a growing area in NLP focused on modeling face-to-face communication). Pre-trained models don't have the necessary components to accept two extra modalities of vision and acoustic. In this paper, we proposed an attachment to BERT and XLNet called Multimodal Adaptation Gate (MAG). MAG allows BERT and XLNet to accept multimodal nonverbal data during fine-tuning. It does so by generating a shift to internal representation of BERT and XLNet; a shift that is conditioned on the visual and acoustic modalities. In our experiments, we study the commonly used CMU-MOSI and CMU-MOSEI datasets for multimodal sentiment analysis. Fine-tuning MAG-BERT and MAG-XLNet significantly boosts the sentiment analysis performance over previous baselines as well as language-only fine-tuning of BERT and XLNet. On the CMU-MOSI dataset, MAG-XLNet achieves human-level multimodal sentiment analysis performance for the first time in the NLP community.