About Me!

I am a Computer Science PhD candidate at University of Toronto, supervised by Babak Taati and Andrea Iaboni. I am also affiliated with KITE - University Health Network as a Research Assistant. Previously, I was a Researcher at Vision & Learning for Autonomous AI (VL4AI) Lab, Monash University.

My research interests broadly span Computer Vision and Deep Learning. I am particularly interested in developing generative models, for 3D human motion analysis using various human sensing technologies, such as video and sensory data, with applications in healthcare.

Education

• Ph.D. in Computer Science, University of Toronto (2021 - Present)
  Thesis: Computer Vision-based human pose estimation, gait assessment, and fall risk prediction in older adults with dementia.

• M.Sc. in Artificial Intelligence, Ferdowsi University of Mashhad (2015 - 2018)
  Thesis: Multi-Stream Human Action Recognition Using Spatiotemporal Saliency Maps

• B.Sc. in Computer Engineering, Ferdowsi University of Mashhad (2010 - 2014)
  Thesis: Object Recognition Using RGB-D Data

Projects

• Generative Model for Human Motion and Pathology Analysis: Designed a generative model based on residual VQ-VAE with transformers and VQ-diffusions for disentangled motion and pathology representation learning and generation, particularly for Parkinson’s Disease.

• EMotionDiffuse-GPT: Multi-modal Motion and Video Generation: A multi-modal project integrating human motion analysis, video interpretation, and emotional/body language insights. Used Vicuna and Llama large language models (LLMs) with advanced diffusion models to generate controllable, emotion-aware motion sequences from video data.

• AMBIENT Project: Using computer vision technology to analyze gait and predict short-term falls risk in older adults with dementia.
  Collaborating with the KITE Research Institute at Toronto Rehabilitation Institute, UHN.

• Interaction-aware Human Pose and Motion Forecasting: In collaboration with Stanford University and Monash University, this project focuses on predicting human motion and interactions in complex environments.
  Partners: Prof. Hamid Rezatofighi, Prof. Ian Reid, Prof. Juan Carlos Niebles, Prof. Silvio Savarese.

• NLP and ML-based Stock Market Performance and Risk Analysis: Developed a machine learning model to assess stock market trends using natural language processing.
  Supervised by Prof. Ehsan Fazl-Ersi, in collaboration with OcularAI Inc.

• Human Physical Demand Recognition for AI Job Analysis: Used body pose estimation techniques to assess physical demand in video-based AI job analysis.
  Collaborated with OcularAI Inc.

Publications

Journals

• 2023: V. Adeli et al., Vision-based Ambient Monitoring of Gait for Dynamic and Short-Term Falls Risk Assessment in People With Dementia, IEEE Journal Of Biomedical And Health Informatics (JBHI).
• 2020: V. Adeli et al., Socially and Contextually Aware Human Motion and Pose Forecasting, IEEE Robotics and Automation Letters (RA-L).
• 2019: V. Adeli et al., A Component-based Video Content Representation for Action Recognition, Image and Vision Computing.

Conferences

• 2024: V. Adeli et al., Benchmarking Skeleton-based Motion Encoder Models for Clinical Applications: Estimating Parkinson’s Disease Severity in Walking Sequences, IEEE International Conference on Automatic Face and Gesture Recognition (FG2024).
• 2024: S. Mehraban, V. Adeli et al., MotionAGFormer: Enhancing 3D Human Pose Estimation with a Transformer-GCNFormer Network, IEEE/CVF Winter Conference on Applications of Computer Vision (WACV2024).
• 2024: I. Klabucar, V. Adeli et al., Evaluating AI Models for Predicting Gait Severity in Parkinson’s Disease, MDM2024.
• 2023: C. Malin-Mayor, V. Adeli et al., Pose2Gait: Extracting Gait Features from Monocular Video, AmI4HC at MICCAI23.
• 2021: V. Adeli et al., TRiPOD: Human Trajectory and Pose Dynamics Forecasting in the Wild, ICCV2021.
• 2020: V. Adeli et al., Socially and Contextually Aware Human Motion and Pose Forecasting, IROS2020.
• 2018: V. Adeli et al., Enhancing Human Action Recognition through Temporal Saliency, International Conference on Pattern Recognition and Artificial Intelligence (ICPRAI 2018), Canada.
• 2018: V. Adeli et al., A New Representation For Human Activity Recognition in Videos Using a Saliency Map, 4th International Conference on New Studies in Computer and IT, Iran.

For a complete list of publications, visit my Google Scholar profile.

Benchmarks and Workshops

• SoMoF: Created SOcial MOtion Forecasting Benchmark (SoMoF), a key resource for human motion forecasting research.
• ICCV2021 Workshop: Organized the 1st Workshop on Human Trajectory and Pose Dynamics Forecasting in the Wild.

R&D Experiences

• Research Assistant, Toronto Rehabilitation Institute (UHN KITE) (2021 - Present)
• Research Assistant, Vision & Learning for Autonomous AI Lab (Monash University) (2019 - 2021)
• Research Associate, Machine Vision Lab, Ferdowsi University of Mashhad (2018 - 2021)
• AI Specialist, OcularAI Inc., Toronto (2018 - 2021)
• Research Assistant, Machine Vision Lab, Ferdowsi University of Mashhad (2015 - 2018)

Honors and Awards

• Outstanding Reviewer, IEEE Conference on Computer Vision and Pattern Recognition (CVPR2022)
• Mount Sinai Hospital Graduate Scholarship in Science and Technology (2022)
• Best Paper Award, 4th International Conference on New Studies in Computer and IT (2018)
• Ranked 1st, Among M.Sc. graduates of AI program, FUM (2018)
• Ranked 2nd, Among all B.Sc. students in Computer Engineering, FUM (2015)

Contact

Feel free to reach out:

• LinkedIn
• GitHub
• Email : vadeli@cs.toronto.edu / vida.adeli@mail.utoronto.ca