MAHDI RAD

Selected publications

Explore the visual story Full list on Google Scholar

3D Pose Estimation

Object position and orientation from images.

  1. BB8: A Scalable, Accurate, Robust to Partial Occlusion Method for Predicting the 3D Poses of Challenging Objects Without Using Depth
  2. A Novel Representation of Parts for Accurate 3D Object Detection and Tracking in Monocular Images
  3. Robust 3D Object Tracking from Monocular Images Using Stable Parts
  4. Making Deep Heatmaps Robust to Partial Occlusions for 3D Object Pose Estimation
See the visual story

Image Normalization

Object detection and pose under challenging lighting.

  1. ALCN: Adaptive Local Contrast Normalization
  2. ALCN: Meta-Learning for Contrast Normalization Applied to Robust 3D Pose Estimation
See the visual story

Hand–Object Interaction

Hands, objects, and their shared 3D motion.

  1. Feature Mapping for Learning Fast and Accurate 3D Pose Inference from Synthetic Images
  2. Domain Transfer for 3D Pose Estimation from Color Images without Manual Annotations
  3. HOnnotate: A Method for 3D Annotation of Hand and Object Poses
  4. Keypoint Transformer: Solving Joint Identification in Challenging Hands and Object Interactions for Accurate 3D Pose Estimation
  5. Hand-4DGS: Feed-Forward 3D Gaussian Splatting for 4D Hand Reconstruction from Egocentric Videos
See the visual story

Room Layout Estimation

Reconstructing the structure of indoor spaces.

  1. General 3D Room Layout from a Single View by Render-and-Compare
  2. MonteFloor: Extending MCTS for Reconstructing Accurate Large-Scale Floor Plans
  3. MCTS with Refinement for Proposals Selection Games in Scene Understanding
See the visual story

Action Recognition

Understanding activities in everyday video.

  1. HoloAssist: An Egocentric Human Interaction Dataset for Interactive AI Assistants in the Real World
  2. EPFL-Smart-Kitchen: An Ego-Exo Multi-Modal Dataset for Challenging Action and Motion Understanding in Video-Language Models
  3. Multi Activity Sequence Alignment via Implicit Clustering
  4. CaSAR: Contact-aware Skeletal Action Recognition
See the visual story

Spatial VLM

Video, language, and spatial reasoning.

  1. Loc3R-VLM: Language-based Localization and 3D Reasoning with Vision-Language Models
  2. Space3D-Bench: Spatial 3D Question Answering Benchmark
  3. Long Time No See: Benchmarking VLMs for Out-of-Sight Spatiotemporal Reasoning in Egocentric Videos
See the visual story

Efficient VLM

More video understanding with fewer tokens.

  1. CoPE-VideoLM: Leveraging Codec Primitives for Efficient Video Language Modeling
  2. AdaptToken: Entropy-based Adaptive Token Selection for MLLM Long Video Understanding
See the visual story