1. Human Pose Estimation 3D Work out List

1.1. Paper Research

REPRESENTATION LEARNING OF VERTEX HEATMAPS FOR 3D HUMAN MESH RECONSTRUCTION FROM MULTI-VIEW IMAGES

Learnable Triangulation of Human Pose

Learning 3D Human Pose Estimation from Dozens of Datasets using a Geometry-Aware Autoencoder to Bridge Between Skeleton Formats

Expressive Body Capture: 3D Hands, Face, and Body from a Single Image

Deep 3D human pose estimation: A review

2D/3D Pose Estimation and Action Recognition using Multitask Deep Learning

ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders

DiffPose: Spatio Temporal Diffusion Model for Video-Based Human Pose Estimation

Diffusion-Based 3D Human Pose Estimation with Multi-Hypothesis Aggregati

1.2. Model Research

1.2.1 Base Architecture

• 기본적으로 diffusion의 네트워크를 따라가기로 했다.

• diffusion의 방식대로 comvis와 hugging face의 stable diffusion 구조를 참조하여 Auto Encoder를 세웠다.
• ‣