Muneeb Ahmed Khan

Postdoctoral Researcher

Vision & Learning Lab, Ulsan National Institute of Science and Technology (UNIST), Korea

I am a postdoctoral researcher in the Vision & Learning Lab at UNIST, working with Prof. Seungryul Baek. I completed my Ph.D. at Sangmyung University in 2025 under Prof. Heemin Park, on efficient and interpretable deep learning for resource-constrained vision.

My research runs along two threads: medical image analysis, and embodied vision for dexterous manipulation and 3D humans. Across both, I care about models that stay reliable when inputs are missing or degraded, and whose decisions can be inspected.

Research

Medical image analysis

Multimodal and interpretable models for clinical imaging: brain-tumor detection and segmentation from MRI, multi-label thoracic disease classification from chest X-rays, and fusion across imaging modalities. Current work targets segmentation that degrades gracefully when MRI sequences are missing.

Embodied vision: hands, humans, robots

Grasp-motion generation for high-DoF robotic hands, hand–object understanding in vision-language models, and 3D human reconstruction and tracking. Our single-shot trajectory-warping approach placed 2nd in the Dexterous Grasp Motion Challenge at the HANDS Workshop, ECCV 2026.

  • Single-shot trajectory warping for grasp motion generation (HANDS 2026 challenge report, ECCV 2026 DexHAND Workshop)
  • HandVQA, fine-grained spatial reasoning about hands in VLMs (CVPR 2026)
  • FlexiAvatar and Multi-THuMBS, 3D Gaussian avatars and multi-person mesh tracking (ECCV 2026)

Earlier work: efficient, interpretable CNNs for traffic-sign recognition and wildfire detection, and ghosting-artifact detection for mobile imaging under two Google Korea (Pixel Camera team) research grants, including the M-GAID dataset (WACV 2025 Workshops).

News

  • Sep 2026Our 2nd-place solution to the Dexterous Grasp Motion Challenge was presented at the HANDS Workshop award session at ECCV 2026 in Malmö. The technical report is accepted at the DexHAND Workshop.
  • Aug 2026Team UVLL HandDex (Vision & Learning Lab, UNIST) placed 2nd in the Dexterous Grasp Motion Challenge, HANDS Workshop at ECCV 2026, with the top easy-track score on the leaderboard.
  • Jul 2026Two papers accepted at ECCV 2026: Multi-THuMBS and FlexiAvatar.
  • Apr 2026HP-ViT, pathology-aware hierarchical transformers for multi-label chest X-ray classification, published in Discover Computing (open access).
  • Mar 2026HandVQA accepted at CVPR 2026.
  • Nov 2025MediFusionNet published in the International Journal of Imaging Systems and Technology.
  • Sep 2025Joined the Vision & Learning Lab at UNIST as a Postdoctoral Researcher.
  • Aug 2025Ph.D. awarded by Sangmyung University.
  • Feb 2025M-GAID, a real-world dataset for ghosting-artifact detection in mobile imaging, presented at the WACV 2025 Workshops; dataset released.

Selected publications

  1. Multi-THuMBS: Multi-person Tracking of 3D Human Meshes Beyond Video Shots

    J. On, M. S. Ali, M. A. Khan, S. Park, I. Moon, H. J. Chang, J. Kim, S. J. Ha, S. Baek

    European Conference on Computer Vision (ECCV), 2026

    arXiv

  2. FlexiAvatar: Unified 3D Gaussian Human Avatars Under Arbitrary Body Visibility

    Y. Y. Tiruneh, M. S. Ali, U. Jeong, M. A. Khan, M. K. C. Sayem, A. Bayramgeldiyev, B. Bhattarai, S. Baek

    European Conference on Computer Vision (ECCV), 2026

    arXiv

  3. 2nd Place Solution to the HANDS 2026 Workshop Challenge – Dexterous Grasp Motion Track: Single-Shot Trajectory Warping for Grasp Motion Generation

    M. A. Khan, W. Kim, S. Kim, M. Munsif, B. Bhattarai, S. Baek

    ECCV 2026 DexHAND Workshop, technical report (non-archival), 2026. 2nd place, Dexterous Grasp Motion Challenge

  4. HandVQA: Diagnosing and Improving Fine-Grained Spatial Reasoning about Hands in Vision-Language Models

    M. K. C. Sayem, M. T. Chowdhury, Y. Y. Tiruneh, M. A. Khan, M. S. Ali, B. Bhattarai, S. Baek

    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

    arXiv

  5. Pathology-Aware Hierarchical Transformers for Multi-Label Thoracic Disease Classification Using Chest X-Rays

    M. A. Khan, H. Park, K. Zagarzusem, S. Paek

    Discover Computing 29, 230, 2026 (open access)

    paper

  6. MediFusionNet: A Novel Architecture for Multimodal Medical Image Analysis

    M. A. Khan, H. Park, D. Yamkhin, S. Paek

    International Journal of Imaging Systems and Technology 35, 2025

    paper

  7. Traffic Sign Recognition Under Visual Perturbations: Shadows, Light Patches, and Simulated Obstructions

    M. A. Khan, Y. Choi, J. Eum, H. Park

    IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW, SAIAD), 2025

    paper

  8. M-GAID: A Real-World Dataset for Ghosting Artifact Detection and Removal in Mobile Imaging

    M. A. Khan, H. Kim, J. Eum, Y. Myung, Y. Choi, H. Park

    IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW, WACI), pp. 1502–1511, 2025

    paper dataset

All 30 publications (19 first-author). Thread colors: medical imaging, embodied vision, other.

In progress

  1. EviMamba

    An evidence-gated selective state-space model for brain-tumor segmentation that remains reliable when MRI sequences are missing.

    First author. Under review, IEEE Transactions on Medical Imaging.

  2. Dexterous Grasp Generation in the Age of Foundation Models

    A survey of grasp generation from parallel-jaw grippers to anthropomorphic hands, organized around analytical, generative, reinforcement-learning and foundation-model paradigms, with a quantitative analysis of why reported success rates are not mutually comparable.

    With Prof. Seungryul Baek. Manuscript in preparation.

  3. RSNA 2026 Knee Abnormality Detection

    Competition entry: ensembles of 2.5D attention-based multiple-instance models and 3D networks for multi-label abnormality detection in knee MRI.

    Kaggle, ongoing.

Education and positions

  • 2025–Postdoctoral Researcher, Vision & Learning Lab, UNISTAdvisor: Prof. Seungryul Baek
  • 2021–2025Ph.D. in Software (Deep Learning & Computer Vision), Sangmyung University, KoreaAdvisor: Prof. Heemin Park. Dissertation: Efficient and Interpretable Deep Learning Frameworks for Resource-Constrained and Time-Sensitive Vision Applications
  • 2019–2021Lecturer, Institute of Southern Punjab, Multan, Pakistan
  • 2019M.S. in Information Technology, National University of Sciences and Technology (NUST), IslamabadThesis: Prediction-Based Target Tracking in Wireless Sensor Networks
  • 2014B.Sc. in Computer Engineering, COMSATS Institute of Information Technology, Lahore

Funding and honors

  • 20262nd place, Dexterous Grasp Motion Challenge, HANDS Workshop at ECCV 2026 (team UVLL HandDex)
  • 2024–2025Google Korea Research Grant, Pixel Camera team: post-processing methods for artifact removal
  • 2023–2024Google Korea Research Grant, Pixel Camera team: objective quality metrics for ghosting artifacts
  • 2023Best Paper Award, Korea Computer Congress (FireXplainer)
  • 2021–2025Professor Scholarship for Ph.D. studies, Sangmyung University
  • 2022–2023DURE Scholarship for international collaboration with Mongolia
  • 2021–2022Teaching Assistant Scholarship, Department of Software, Sangmyung University