Vida Adeli

Splat2Mesh: From Gaussian Splats to Structured and Editable 3D Scenes

arXiv

Splat2Mesh transforms Gaussian splat scenes into structured, object-centric 3D assets with editable, textured meshes.

PUPPETEER: Object-Grounded Posture-Aware Co-Speech Gesture Generation via Casual Latent Diffusion

Vida Adeli*, Soroush Mehraban*, Jacob Rommann, Harrison Sanborn, Cole Clifford, Babak Taati

arXiv

Puppeteer generates speech-aligned 3D gestures that adapt to body posture and surrounding objects using causal latent diffusion, enabled by SceneGes, a new dataset for object-grounded co-speech gestures.

PickStyle: Video-to-Video Style Transfer with Context-Style Adapters

Soroush Mehraban*, Vida Adeli*, Jacob Rommann, Babak Taati, Kyryl Truskovskyi

* Equal contribution

European Conference on Computer Vision Workshops (Oral), 2026

PickStyle is a diffusion-based video style transfer framework that preserves video context while applying a target visual style. It uses low-rank style adapters and synthetic clip augmentation from paired images for training, and introduces Context-Style Classifier-Free Guidance (CS-CFG) to independently control content and style, achieving temporally consistent and style-faithful video results.

GAITGen: Disentangled Motion-Pathology Impaired Gait Generative Model

Vida Adeli, Soroush Mehraban, Majid Mirmehdi, Alan Whone, Benjamin Filtjens, Amirhossein Dadashzadeh, Alfonso Fasano, Andrea Iaboni, Babak Taati

Winter Conference on Applications of Computer Vision (WACV), 2026

GAITGen is a generative framework that synthesizes realistic gait sequences conditioned on Parkinson’s severity. Using a Conditional Residual VQ-VAE and tailored Transformers, it disentangles motion and pathology features to produce clinically meaningful gait data. GAITGen enhances dataset diversity and improves performance in parkinsonian gait analysis tasks.

PainControl: Identity-Preserving Pain Expression Transfer with Generative Diffusion Models

Yasamin Zarghami, Muhammad Muzammil, Vida Adeli, Hailey Reimer, Thomas Hadjistavropoulos, Babak Taati

BioMedical Engineering OnLine, 2026

PainControl is a facial landmark-guided method for identity-preserving pain expression transfer.

CARE-PD: A Multi-Site Anonymized Clinical Dataset for Parkinson's Disease Gait Assessment

Vida Adeli, Ivan Klabucar, Javad Rajabi, Benjamin Filtjens, Soroush Mehraban, et al.

Neural Information Processing Systems (NeurIPS), 2025

CARE-PD is the largest publicly available archive of 3D mesh gait data for Parkinson’s Disease, collected across 9 cohorts from 8 clinical centers. It provides standardized, anonymized SMPL representations and benchmark protocols for clinical motion analysis on PD.

Benchmarking Skeleton-based Motion Encoder Models for Clinical Applications

Vida Adeli, Soroush Mehraban, Irene Ballester, Yasamin Zarghami, Andrea Sabo, Andrea Iaboni, Babak Taati

IEEE International Conference on Automatic Face and Gesture Recognition (FG), 2024

Evaluating recent motion encoders for the task of parkinsonism severity estimation (UPDRS III gait)

MotionAGFormer: Enhancing 3D Pose Estimation with a Transformer-GCNFormer Network

Soroush Mehraban, Vida Adeli, Babak Taati

Winter Conference on Applications of Computer Vision (WACV), 2024

Estimating 3D locations of 17 main joints from a monocular video.

Ambient Monitoring of Gait for Dynamic and Short-Term Falls Risk Assessment in People With Dementia

Vida Adeli, Soroush Mehraban, Navid Korhani, Andrea Sabo, Sina Mehdizadeh, Alastair Flint, Avril Mansfield, Andrea Iaboni, Babak Taati

IEEE journal of biomedical and health informatics (JBHI), 2023

A machine learning system using ambient gait monitoring to predict short-term fall risk in people with dementia.

TRiPOD: Human Trajectory and Pose Dynamics Forecasting in the Wild

Vida Adeli, Mahsa Ehsanpour, Ian Reid, Juan Carlos Niebles, Silvio Savarese, Ehsan Adeli, Hamid Rezatofighi

IEEE/CVF International Conference on Computer Vision, (ICCV 2021)

TRiPOD jointly forecasts human body pose dynamics and global trajectories in the wild by modelling human-human and human-object interactions.

Socially and Contextually Aware Human Motion and Pose Forecasting

Vida Adeli, Ehsan Adeli, Ian Reid, Juan Carlos Niebles, Hamid Rezatofighi

IEEE Robotics and Automation Letters (RA-L) and IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020

A unified end-to-end model that predicts both human global motion (trajectory) and detailed body pose jointly, using social and scene context to improve forecasting.

A Component-Based Video Content Representation for Action Recognition

Vida Adeli, Ehsan Fazl-Ersi, Ahad Harati

Image and Vision Computing, 2019

Proposes a component-based video content representation for human action recognition, decomposing videos into meaningful parts to improve classification in complex scenes.