Pose Embeddings

Pose embeddings represent the posture or configuration of an object or person in a compact, numerical form, aiming to capture essential information for tasks like object recognition, action recognition, and avatar generation. Current research focuses on developing robust and efficient methods for learning these embeddings, often employing attention-based dual-encoder architectures, diffusion models, or keypoint-based hybrid representations to disentangle pose from object identity and improve generalization across different objects and viewpoints. This work has significant implications for various applications, including improved object recognition in challenging conditions, more realistic human avatar animation, and enhanced video action understanding.

Papers

November 13, 2024

Generalized Pose Space Embeddings for Training In-the-Wild using Anaylis-by-Synthesis
Dominik Borer, Jakob Buhmann, Martin Guay
Synthetic Data Human Pose Neural Rendering Skeleton Representation Automatic Synthesis Pose Embeddings

October 11, 2024

HpEIS: Learning Hand Pose Embeddings for Multimedia Interactive Systems
Songpei Xu, Xuri Ge, Chaitanya Kaul, Roderick Murray-Smith
Variational Autoencoder Hand Object Interaction Hand Pose Estimation Hand Pose Virtual Sensor Interactive System Pose Embeddings

June 14, 2024

Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses
Seungwoo Yoo, Juil Koo, Kyeongmin Yeo, Minhyuk Sung
Human Pose Deformable Object Object Pose Pose Transfer Pose Embeddings

March 1, 2024

Dual Pose-invariant Embeddings: Learning Category and Object-specific Discriminative Representations for Recognition and Retrieval
Rohan Sarkar, Avinash Kak
App to App Retrieval Recognition Rate Object Embeddings Pose Invariant Pose Embeddings

August 7, 2023

ViLP: Knowledge Exploration using Vision, Language, and Pose Embeddings for Video Action Recognition
Soumyabrata Chaudhuri, Saumik Bhattacharya
Vision Paper Multi Modal Learning Stable Pose Video Action Recognition Video Pretraining Knowledge Exploration Pose Embeddings

April 25, 2023

PoseVocab: Learning Joint-structured Pose Embeddings for Human Avatar Modeling
Zhe Li, Zerong Zheng, Yuxiao Liu, Boyao Zhou, Yebin Liu
Human Pose Stable Pose Global Pose Avatar Modeling Pose Embeddings

April 19, 2022

ActAR: Actor-Driven Pose Embeddings for Video Action Recognition
Soufiane Lamghari, Guillaume-Alexandre Bilodeau, Nicolas Saunier
Action Recognition Human Action Recognition Action Representation Video Action Recognition Pose Embeddings Infrared Video

Pose Embeddings

Papers

Generalized Pose Space Embeddings for Training In-the-Wild using Anaylis-by-Synthesis

HpEIS: Learning Hand Pose Embeddings for Multimedia Interactive Systems

Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses

Dual Pose-invariant Embeddings: Learning Category and Object-specific Discriminative Representations for Recognition and Retrieval

ViLP: Knowledge Exploration using Vision, Language, and Pose Embeddings for Video Action Recognition

PoseVocab: Learning Joint-structured Pose Embeddings for Human Avatar Modeling

ActAR: Actor-Driven Pose Embeddings for Video Action Recognition